The research, published in Frontiers in Sustainability by Vita Jagrič of the University of Maribor, Dijana Oreški of the University of Zagreb, and Polona Tominc and Maja Rožman of the University of Maribor, set out to answer a deceptively simple question: does structured ESG risk governance actually relate to favorable organizational outcomes in non-financial firms? The answer, based on a survey of 114 medium-sized and large companies, is a qualified yes, and the strength of the relationship depends on which dimension of governance a firm prioritizes.

The study rests on a theoretical foundation drawn from institutional theory, stakeholder theory, and knowledge-transfer perspectives. Institutional theory explains why practices diffuse across organizational fields: coercive pressures from regulation, normative pressures from professional standards, and mimetic pressures through which uncertain firms emulate organizations perceived as more advanced. Banks, operating under detailed European Banking Authority guidelines on ESG risk management, have institutionalized structured approaches earlier and more comprehensively than firms in less-regulated sectors. The researchers argue that non-financial firms facing growing but less codified ESG expectations may emulate these mature practices to reduce uncertainty and enhance legitimacy, adapting them selectively rather than replicating them wholesale.

To ground this argument in practice, the team conducted a structured document-based benchmark of two major Slovenian banks, Nova Ljubljanska Banka and OTP Banka, reviewing publicly available sustainability strategies, Pillar III disclosures, governance information, and external ESG assessments published between 2021 and 2025. The benchmark revealed how banks treat ESG risks not as a new risk category but as an aggravating factor for risks they already manage. NLB, for example, monitors climate risk factors through traditional categories such as credit, market, liquidity, operational, and reputational risk, using transmission channels that link environmental drivers to financial exposures. The bank maintains a four-level sustainability governance structure, integrates environmental and social risks into its risk appetite statement, runs its own ESG stress-testing concept, and operates a three-level internal control system spanning business units, the risk management function, compliance, and independent internal audit. In 2024, Sustainalytics rated NLB’s ESG risk management as strong, scoring 96.2 out of 100.

These banking practices served as contextual illustrations rather than empirical comparison groups. The actual hypothesis testing came from a survey administered online between April and May 2026 to 350 randomly selected medium-sized and large non-financial firms in Slovenia and Croatia, identified from national registers of annual financial statements. After excluding incomplete responses, 114 valid responses remained, a response rate of 32.6 percent. Respondents were senior informants with direct knowledge of their firm’s risk management and ESG practices, including risk managers, chief financial officers, and members of senior management. The sample skewed toward manufacturing, which accounted for 38.6 percent of firms, followed by trade at 24.6 percent, with 58.8 percent classified as medium-sized enterprises under EU criteria.

The researchers conceptualized ESG risk governance through three distinct dimensions, each measured with four survey items on a five-point Likert scale. Environmental risk governance and integration captured whether environmental risks are formally incorporated into enterprise risk management frameworks, whether structured identification and assessment procedures exist, whether indicators monitor exposure, and whether assessments influence strategic planning. Social risk governance and integration addressed the integration of social and stakeholder-related risks into formal processes, structured policies, senior management review, and reporting mechanisms. ESG risk governance and oversight measured the clarity of assigned ESG responsibilities, board or top management oversight, embedding of ESG risks within internal control and compliance systems, and adherence to defined governance procedures. The outcome construct, perceived ESG-related organizational outcomes, reflected managers’ assessments of effective integration, organizational stability, long-term strategic resilience, and overall performance.

The relationships were assessed using partial least squares structural equation modeling, implemented in WarpPLS software. The measurement model demonstrated strong psychometric properties, with all indicator loadings exceeding 0.700, Cronbach’s alpha values ranging from 0.841 to 0.879, composite reliability between 0.893 and 0.914, and average variance extracted from 0.661 to 0.724, all above recommended thresholds. Discriminant validity was confirmed through heterotrait-monotrait ratios below 0.85 and the Fornell-Larcker criterion. Common method bias, a concern given the single-respondent design, was assessed using full-collinearity variance inflation factors, which ranged from 1.731 to 1.928, well below the conservative threshold of 3.3, and Harman’s single-factor test, in which the first unrotated factor accounted for 38.637 percent of variance, below the 50 percent criterion.

The structural results supported all three hypothesized relationships and revealed a clear ordering. ESG risk governance and oversight showed the strongest association with perceived organizational outcomes, with a standardized path coefficient of 0.402 and a large effect size of 0.372. Environmental risk governance and integration followed at 0.360 with an effect size of 0.241, while social risk governance and integration showed a comparatively moderate but still significant association at 0.289 with an effect size of 0.164. Together, the three dimensions explained 58.6 percent of the variance in perceived ESG-related organizational outcomes, with a predictive relevance coefficient of 0.421 confirming adequate predictive capability.

Robustness checks strengthened confidence in these findings. Adding firm size, country, and industry controls produced only minor changes in path coefficients, with none of the control variables reaching statistical significance, and explained variance increasing modestly from 0.586 to 0.618. A multi-group comparison found no statistically significant differences between Croatian and Slovenian subsamples, with oversight remaining the strongest predictor in both countries. An endogeneity assessment using stochastic instrumental variables detected no statistical evidence that the estimated relationships were materially affected by omitted variables, simultaneity, or reverse causality, though the researchers caution that this diagnostic cannot establish causal identification in a cross-sectional design.

The practical implications are direct. Non-financial firms seeking stronger ESG outcomes should prioritize board-level oversight, clearly assigned ESG responsibilities, formal integration of ESG risks into enterprise risk management, and systematic monitoring and reporting. The researchers suggest establishing dedicated ESG oversight committees, integrating ESG risk metrics into executive performance evaluation, and embedding ESG considerations into strategic planning and capital allocation. Governance, in this view, is not merely one ESG pillar among others but the structural architecture through which environmental and social risk practices become coordinated and operationalized. The findings also carry implications for policymakers, suggesting that regulatory emphasis on governance structures can catalyze more consistent and strategically embedded ESG implementation across industries, reducing fragmentation in sustainability practices and improving comparability across firms and sectors.

The study acknowledges limitations that temper interpretation. The cross-sectional, single-respondent design limits causal inference, and the outcome construct reflects managerial perceptions rather than externally verified ESG ratings or objective performance indicators. The sample of 114 firms, while adequate for the model estimated, is confined to two Central and Eastern European economies sharing a comparable post-transition institutional heritage, and generalization to other contexts should proceed with caution. Future research, the authors suggest, should employ longitudinal and multi-source data, combine survey responses with archival financial data and third-party ESG ratings, and examine whether the relative importance of environmental, social, and governance dimensions varies across institutional settings. For now, the message for managers is clear: fragmented, compliance-driven sustainability initiatives are unlikely to deliver the organizational benefits that structured, governance-anchored ESG risk management appears to support.

Subject of Research: ESG risk governance and perceived organizational outcomes in non-financial firms in Slovenia and Croatia

Article Title: From banking-sector benchmarks to non-financial firms: ESG risk governance and perceived organizational outcomes in Slovenia and Croatia

Article References: Jagrič, V., Oreški, D., Tominc, P., & Rožman, M. (2026). From banking-sector benchmarks to non-financial firms: ESG risk governance and perceived organizational outcomes in Slovenia and Croatia. Frontiers in Sustainability, 7, Article 1965019. https://doi.org/10.3389/frsus.2026.1965019

Image Credits: AI Generated

DOI: 10.3389/frsus.2026.1965019

Keywords: ESG, risk governance, enterprise risk management, Slovenia, Croatia, PLS-SEM, sustainability, banking-sector, benchmarks, non-financial, firms, risk

The editorial, written by Bruno César Feltes, Raul Santos and Hélio Veríssimo of the Center for Molecular Biology and Biotechnology, draws together a collection of seven research articles that probe the hidden fragility of computational biology. Their conclusion is uncomfortable reading for anyone who assumed that running a standard pipeline guarantees a standard answer. Even when two tools claim to perform the same task, the biological output can differ in term specificity, pathway prioritization and stringency, meaning that the choice of software is itself a scientific decision with real consequences.

One of the clearest demonstrations comes from a study that benchmarked twelve widely used over-representation analysis tools against controlled and experimental datasets. The researchers found that the tools produced markedly different biological outputs, differing in the specificity of the terms they reported, the pathways they prioritized and the stringency of their statistical thresholds. Because all twelve tools relied on the same underlying strategy and the same annotation database, the differences could not be attributed to data quality or database version. Conventional performance metrics, the authors argue, are simply inadequate for capturing how interpretable and biologically meaningful the results actually are.

The fragility extends beyond pathway analysis. A separate study examined the detection of SARS-CoV-2 variants from sequencing data and found that performance varied dramatically depending on the sequencing protocol, the variant-calling software and the primer schemes used to enrich the samples. In other words, the same virus, sequenced from the same patient, could yield different variant profiles depending on the experimental and computational choices made along the way. The authors conclude that benchmarks must be designed around the full experimental context, not just the algorithm.

Machine learning introduces its own set of pitfalls. A workflow designed to classify antimicrobial peptides, a notoriously difficult task because of the extreme diversity of these molecules, showed how protein-specific factors such as homologous sequences, structural features and evolutionary signals can systematically bias a classifier’s apparent performance. The authors present a case study in which a model trained on antimicrobial peptides achieved impressive cross-validation scores but failed to generalize to real-world data, a cautionary tale for anyone applying deep learning to biological sequences.

Software engineering itself is part of the solution. One contribution approaches bioinformatics from the perspective of workflow management, arguing that the field should borrow practices from large-scale software development, including version control, containerization, automated testing and continuous integration. The authors contend that treating pipelines as engineering artifacts, subject to the same rigor as laboratory instruments, is essential for producing results that can be reproduced years later, even as the underlying tools and databases continue to evolve.

Accessibility is another recurring theme. A theoretical framework for genomics and transcriptomics education argues that newcomers to computational biology are often overwhelmed by the sheer number of tools, databases and opinions available online. The authors propose organizing the field around the core concepts that underlie all workflows, rather than around individual software packages, so that the next generation of researchers can evaluate methods critically instead of simply following the most popular tutorial.

Perhaps the most sobering message comes from a study of antimicrobial peptide classification that examined how training data composition affects model behavior. The authors showed that the apparent performance of a classifier can be inflated by homologous sequences that leak between training and test sets, a phenomenon that is especially dangerous in biological applications where the ultimate goal is to discover genuinely novel molecules. Their findings suggest that many published benchmarks in computational biology may overstate the real-world capabilities of the models they evaluate.

Taken together, the seven studies collected in this special issue paint a picture of a field at an inflection point. The tools of bioinformatics are more powerful than ever, but the editorial argues that their outputs are only as trustworthy as the experimental context, the training data and the engineering practices behind them. As computational biology continues to expand into clinical diagnostics, drug discovery and public health, the authors contend that rigorous, context-aware benchmarking is no longer optional. It is, they conclude, the foundation upon which the reliability of the entire discipline rests.

One way to understand why benchmarking has become so central to bioinformatics is to consider how the field’s growth has changed the nature of methodological choice. A decade ago, a laboratory might have had a handful of options for aligning reads or annotating functions; today, the same task can be approached with dozens of tools, each with its own parameter space, dependencies and assumptions. The editorial frames this not as an embarrassment of riches but as a structural risk: when analytical choices multiply faster than the evidence base for choosing among them, results become contingent on decisions that are rarely reported, let alone justified. Benchmarking, in this view, is not a one-off exercise but an ongoing practice of mapping which methods behave predictably under which conditions.

The editorial is careful to distinguish benchmarking from a simple ranking exercise. A tool that achieves the highest score on a particular metric under one set of conditions may behave quite differently when dataset composition, input size, sequencing characteristics or representation strategies change. This is why the editors argue that meaningful benchmarking must be anchored to the biological question rather than to abstract performance figures. Two outputs can both be technically acceptable while differing substantially in their biological informativeness and interpretability, which means that the value of a method cannot be assessed in isolation from the research scenario in which it will be deployed. The practical implication is that benchmarking studies should define, and communicate, the domains of application within which their conclusions hold.

Several of the collected articles illustrate this principle from different angles. The study of tumor-only somatic variant calling, for instance, introduced a pipeline called Synth4bench that generates controlled synthetic datasets, allowing the authors to evaluate five widely used callers under conditions they could systematically vary. The finding that indel detection and low-allele-frequency variants remained difficult for all callers, while overall performance diverged with sequencing parameters and algorithmic design, reinforces a recurring lesson: no single tool dominates across all conditions, so matching the algorithm to the analytical objective and the sequencing configuration is itself a methodological decision that deserves explicit justification.

The contribution on reproducibility approaches the same problem from the software-engineering side. Rather than adapting established software engineering principles wholesale, the authors translate them into the specific vocabulary of workflow management systems, arguing for explicit task interfaces, a separation between the computational core of a pipeline and the components that adapt it to particular datasets, systematic validation, comprehensive testing and continuous integration. These practices are connected to FAIR-oriented guidelines for research software, which emphasize that software should be findable, accessible, interoperable and reusable in the same sense that research data are. The underlying argument is that reproducibility is not achieved at the moment of publication but engineered into the pipeline from the start, so that results can be regenerated years later even as dependencies and databases evolve.

Usability receives attention as a dimension of methodological quality that is easy to overlook. The study of novice genomic researchers combined an analysis of the literature with semi-structured interviews, and the obstacles that emerged were strikingly mundane: inadequate documentation, limited accessible guidance, installation and dependency difficulties, and inconsistent error or progress reporting. The authors translated these findings into an evaluation rubric for bioinformatics software, making the case that user-centered design is not a cosmetic concern but a determinant of whether analyses are performed correctly. A tool that cannot be installed reliably or that fails silently is, in practice, a tool whose results cannot be trusted, regardless of the sophistication of its underlying algorithms.

The educational contribution in the collection takes a complementary approach by organizing genomics and transcriptomics around their data products, analytical goals and recurring modular workflow steps rather than around individual packages. This structure is intended to give early-stage researchers a theoretical foundation that remains stable as specific tools rise and fall in popularity, which addresses the dual problem the editors identify: newcomers struggle to navigate the sea of available options, while veterans tend to fall back on the tools they already know, even as better-suited alternatives emerge.

Underlying all seven articles is a shared recognition that methodological complexity now poses barriers that scale with the field’s accessibility. As bioinformatics becomes a routine component of biological research across laboratories with widely varying computational expertise, the consequences of opaque parameterization, undocumented workflows and unvalidated defaults fall on an increasingly broad community of users. The editorial’s framing of benchmarking as a means of defining behavior, limitations and appropriate domains of application, rather than as a competition among tools, offers a vocabulary for addressing this: the goal is not to crown winners but to make the conditions under which each method performs reliably explicit, documented and testable, so that methodological choices can be defended on evidence rather than habit.

Subject of Research: Benchmarking practices and reproducibility of bioinformatics tools and workflows

Article Title: Editorial: Benchmarks and good practices in bioinformatics

Article References: Feltes, B. C., Santos, R. A. C. D., & Verli, H. (2026). Editorial: Benchmarks and good practices in bioinformatics. Frontiers in Bioinformatics, 6, Article 2005704. https://doi.org/10.3389/fbinf.2026.2005704

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.2005704

Keywords: bioinformatics, benchmarking, reproducibility, computational biology, machine learning, genomics, workflows, Frontiers in Bioinformatics, Editorial, Benchmarks, good, practices

The study, conducted by Damian Machlanski of the University of Edinburgh, Spyridon Samothrakis of the University of Essex, and Paul Clarke of the University of Essex, tackles a puzzle at the heart of what statisticians call treatment effect estimation. Using Rubin’s potential outcomes framework, the goal is to estimate the difference between what happened to people who received a treatment and what would have happened to those same people had they not received it. The average of these differences across a population is the Average Treatment Effect, or ATE, while the conditional version, CATE, describes how the effect varies across subgroups. The catch is that one of the two potential outcomes for every individual is always counterfactual—it never happened and can never be observed. Unlike predicting house prices or classifying images, where a validation set reveals exactly how wrong a model is, causal models are graded on answers that are fundamentally hidden.

This hiding is what makes hyperparameter tuning so treacherous in the causal setting. In standard supervised machine learning, everyone agrees on the metric: minimize mean squared error on held-out data. The best-fitting model wins. In causal inference, what researchers actually care about is the error on the difference between potential outcomes, quantified by metrics like the Precision in Estimation of Heterogeneous Effects (PEHE). But because counterfactuals are unobservable, only the observable Mean Squared Error on factual outcomes is available in real datasets. The team formalizes this tension by distinguishing the observable metric, which can be computed, from the potential metric, which is ideal but inaccessible. When treatment assignment is random, as in a randomized controlled trial, the observable error is a decent stand-in for the potential one. Observational data, however, is riddled with covariate shift—the characteristics of treated and untreated individuals differ systematically—and the proxy can fail spectacularly.

To demonstrate the problem vividly, the authors construct a deceptively simple toy example: a one-dimensional input, a sinusoidal data-generating process, and a treated group with missing regions of data. Three candidate models emerge. The first fits the control arm beautifully but misses the trend in the treated arm, achieving the best observable MSE while producing terrible causal estimates. The second captures the true trend in both arms despite the data gaps—the solution anyone would want—but its MSE is worse, so conventional model selection would reject it. The third fits the data poorly overall yet, through leakage of benchmark test information, manages to top the PEHE leaderboard. All three illustrate the same disturbing conclusion: in causal estimation, the metric used to pick a model can be actively misleading, and a solution that looks best on paper may be the wrong one.

The centerpiece of the paper is an exhaustive empirical study built on the four classic causal benchmarks: IHDP, derived from the Infant Health and Development Program; Jobs, combining experimental and observational employment data; Twins, based on twin births in the United States between 1989 and 1991; and News, a high-dimensional dataset of simulated user responses to articles. Crucially, these benchmarks contain ground-truth counterfactuals, which allowed the researchers to compute the otherwise inaccessible potential metrics—not to tune models, which would violate good practice, but to grade how well different observable metrics perform as tuning proxies. The experimental grid spans seven widely used causal estimators, including S-Learners, T-Learners, X-Learners, Doubly Robust estimators, Double Machine Learning, Inverse Propensity Score Weighting, and Causal Forests, crossed with nine base learners ranging from Lasso and Ridge regression to gradient-boosted trees and neural networks.

The headline result is stark. Using an oracle that selects hyperparameters with the ideal potential metrics—the best any data-driven tuning could ever achieve—the probability of reaching state-of-the-art performance jumped from 65 percent to 81 percent for average treatment effect estimation, and from 50 percent to 57 percent for individualized estimation. The comparison held across nearly every estimator and base learner, with typical gains of 8 to 12 percent, and even larger improvements of 10 to 33 percent for learners like Lasso, Decision Trees, Kernel Ridge, and CatBoost. Only the ensemble methods, including Random Forests and Causal Forests, showed robust defaults—a likely explanation for the strong out-of-the-box performance of Causal Forests, whose ensembling structure echoes the resilience of the other ensemble learners.

The implications ripple outward. If swapping hyperparameters alone can push an estimator from the bottom of the league table to state-of-the-art territory, then published benchmarks that compare causal methods may be measuring tuning quality as much as algorithmic merit. The authors point out that different studies use different metrics for hyperparameter optimization, making cross-paper comparisons of estimators inherently fragile. Their conclusion is deliberately provocative: a practitioner who tunes a single model carefully may outperform a rival who searches over many models without tuning. In other words, the casual assumption that newer, more sophisticated neural estimators inherently beat older regression-based approaches may simply reflect differences in how diligently their knobs were turned.

Yet the study also offers a measure of hope. Not every observable metric fails equally. When the team evaluated several practical tuning proxies—including factual MSE, plugin validation that synthesizes pseudo-ground-truth effects with a second estimator, k-nearest-neighbour matching, and the R-Score derived from the R-Learner framework—they found that some came remarkably close to oracle performance. Matching-based plug-in metrics in particular showed low variability, with a 95 percent confidence interval of roughly 0.01 on the aggregated probability scale, suggesting that well-chosen proxies can put excellent hyperparameters within reach. Others fared badly: the R-Score behaved erratically, diverging from recent findings, and plain MSE, while on average close to the oracle, produced inconsistent results across scenarios. The choice of metric, the authors conclude, can seriously distort final estimation performance, and there is currently no metric uniformly capable of identifying the best causal model.

The researchers are candid about the limits of their work. The analysis rests on the standard assumptions of causal inference—no interference between units, no hidden confounders, and positive probability of treatment for everyone—and the search spaces, while broad, had to be trimmed for computational feasibility. Still, their practical recommendation is strikingly modest: until more robust observable metrics are developed, practitioners may be better served by highly interpretable models such as linear regression or decision trees, whose hyperparameters domain experts can tune manually and thereby embed genuine prior knowledge. Perhaps most provocatively, they argue that the field does not need new estimators at all. Meaningful progress, they contend, will come from a deeper understanding of causal model selection and hyperparameter tuning—the same realization that transformed general machine learning a decade ago—potentially aided by standardized public benchmarking platforms that tune every causal estimator equally thoroughly, so that claimed breakthroughs can finally be credited to the methods themselves rather than to lucky settings.

Subject of Research: Hyperparameter tuning and model selection metrics for machine learning-based causal effect estimation

Article Title: The Challenges of Hyperparameter Tuning for Accurate Causal Effect Estimation

Article References: Machlanski, D., Samothrakis, S., & Clarke, P. (2026). The Challenges of Hyperparameter Tuning for Accurate Causal Effect Estimation. Machine Learning, 115(11), Article 249. https://doi.org/10.1007/s10994-026-07140-9

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07140-9

Keywords: causal inference, hyperparameter tuning, treatment effect estimation, CATE, model selection, machine learning, benchmarks, evaluation metrics, double machine learning, observational data, counterfactuals, Challenges

The question sounds trivial until you appreciate how crowded the field has become. Deep learning predictors for RNA 3D structure have proliferated in recent years, inspired by the revolution that AlphaFold brought to protein modeling. Yet unlike protein science, where community-wide CASP experiments established rigorous standards, RNA structure prediction has lacked systematic, controlled benchmarking. Each new method reports its own numbers on its own test cases, making it nearly impossible for a working scientist to know which predictor, run under which conditions, will produce the most accurate model for a fresh RNA sequence. ROMET-RNA, described as an automated benchmarking framework and meta-selection toolkit, was built to close that gap.

The scale of the benchmark is notable. The team ran eleven state-of-the-art RNA structure predictors through a single standardized workflow, eliminating the apples-to-oranges comparisons that plague the literature. Crucially, the researchers also varied something that most benchmarks ignore: the multiple sequence alignment, or MSA, that deep learning models consume as input. These alignments, built from families of evolutionarily related sequences, carry coevolutionary signals — positions that mutate together across evolution because they touch or interact in the folded structure. The models read these signals to infer which nucleotides pair and pack. ROMET-RNA evaluated three complementary MSA pipelines: rMSA, RNAcmap3 MARS, and Infernal RNAcentral, each generating alignments by different strategies.

Model selection in ROMET-RNA rests on a consensus scoring strategy that combines two complementary judges. The first is the TM-scoreRNA, a structural similarity metric that measures how closely a predicted model matches a reference structure in three-dimensional space, rewarding correct global topology rather than just local base-pairing. The second is lociPARSE, a deep learning derived statistical potential that estimates the physical plausibility of a model without needing the true structure at all. By fusing structural similarity with learned physical plausibility, the consensus score aims to pick models that are both geometrically accurate and energetically sensible — a combination that either metric alone can miss.

The benchmark itself spans four non-redundant test sets comprising 205 RNAs, drawn from CASP15 targets, RNA Puzzles challenges, and recent structures deposited in the Protein Data Bank. These communities represent the closest thing RNA structural biology has to public competitions, where labs worldwide attempt the same folding problems blind. Across all four test sets, the results show that ROMET-RNA consistently selects more accurate models than any individual predictor would achieve on its own. The meta-selection, in other words, genuinely beats the best single method — the kind of result that benchmarking studies rarely deliver.

Skeptics of AI benchmarks will appreciate the team’s caution on one point: training-test overlap. Deep learning predictors are trained on known structures, and if a benchmark test set contains RNAs similar to training examples, performance can be inflated by memorization rather than genuine inference. The researchers explicitly confirmed that ROMET-RNA’s advantage is not an artefact of such overlap, strengthening the claim that the toolkit’s consensus strategy reflects real predictive power rather than statistical luck.

Beyond the headline result, the analysis yields several technical insights with immediate practical value. Coevolutionary information, the team found, substantially improves prediction accuracy — confirming that the evolutionary signals embedded in sequence families remain the richest source of structural information available to current models. But the optimal MSA strategy is method specific: a predictor that performs best with one alignment pipeline may falter with another, meaning there is no universal recipe for preparing model inputs. Perhaps most intriguingly, MSA depth — the number of sequences in the alignment — exhibits a nonlinear relationship with performance. More sequences do not always mean better predictions; there appears to be a sweet spot beyond which additional depth yields diminishing or even negative returns.

That nonlinear depth finding matters for anyone running these models at scale. Generating deep alignments is computationally expensive, and the result suggests that simply maximizing alignment size is a wasteful strategy. Instead, practitioners may need to tune alignment depth per target and per predictor, a task that ROMET-RNA’s standardized framework makes tractable by letting users systematically vary inputs and compare outputs under controlled conditions. The supplementary materials include input-sensitivity analyses, statistical comparisons, and worked examples designed to guide such tuning.

The implications reach across RNA structural biology and genomics. Accurate tertiary structure prediction is essential for understanding RNA function, from riboswitches that sense metabolites to long non-coding RNAs whose mechanisms remain mysterious. As the volume of newly sequenced RNA families explodes and experimental structure determination lags far behind, reliable computational prediction becomes the bottleneck for translating sequence data into mechanistic insight. A tool that reliably picks the best available model — and tells you which predictor and alignment strategy to use — could accelerate drug design, ribozyme engineering, and the annotation of the vast non-coding transcriptome.

ROMET-RNA is released as an open-source standardized framework, meaning the benchmarking workflow, the consensus scoring, and the toolkit are available to the community rather than locked behind proprietary walls. The study, led by Chenjie Feng and Tian Zhang as joint first authors with Xin Gao and Renmin Han as corresponding authors, was published in Genome Biology on 15 October 2026. For a field racing to apply deep learning to RNA folding, the message is clear: the era of trusting a single predictor is over, and the future belongs to systems that benchmark broadly, vary inputs deliberately, and select models by consensus.

Subject of Research: Computational biology

Article Title: ROMET-RNA benchmarks and selects RNA 3D structures using multiple predictors and alignment strategies

Article References: Feng, C., Zhang, T., He, W., Liu, Z., Xu, X., Zha, C., Gao, X., & Han, R. (2026). ROMET-RNA benchmarks and selects RNA 3D structures using multiple predictors and alignment strategies. Genome Biology. https://doi.org/10.1186/s13059-026-04308-3

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04308-3

Keywords: RNA structure prediction, ROMET-RNA, deep learning, multiple sequence alignment, Genome Biology, structural biology, benchmarking, coevolution, RNA folding, machine learning, benchmarks, selects

The stakes are far from academic. Practitioners in industry, healthcare, and infrastructure security rely on published benchmark comparisons to choose which detection systems to deploy. If the evidence that sophisticated neural networks beat simpler methods is distorted by flawed evaluation practices, organizations may be investing in expensive, complex models that offer no real advantage. The KTH team’s investigation reads less like a routine review and more like a forensic audit of a field’s collective assumptions, and its findings echo a broader reproducibility crisis that has been gathering force across machine learning research in recent years.

The first and most pervasive problem the researchers identified is data leakage, which occurs when information from the test set—data that should be invisible during evaluation—contaminates the assessment process. In time series anomaly detection, models produce continuous anomaly scores that must be converted into binary predictions via a threshold. The most damaging practice, found in seven of the 18 papers examined, involves selecting that threshold by trying many values on the test set and keeping whichever yields the best score. In any real deployment, no such oracle exists; the threshold must be chosen without knowing the answers. A second variant tunes the threshold so that a fixed percentage of test predictions are flagged as anomalous, often matching the known anomaly rate of the test set—still a form of test-set knowledge that would be unavailable in practice.

Even more striking is the widespread use of the point-adjust algorithm, a post-hoc evaluation trick in which if a model detects at least one anomalous point within a continuous segment of anomalies, the entire segment is credited as correctly detected. Six of the examined papers used this method directly. The authors illustrate its distorting power with a simple example: original predictions scoring 0.5 precision, 0.2 recall, and 0.286 F1 score jump to 0.8 across all three metrics after point adjustment. Previous work has shown that under point adjustment, random guessing can compete with state-of-the-art methods. Because the adjustment compares predictions to ground-truth labels before scoring, it is, in the researchers’ view, a clear case of test-set leakage. In total, eleven of the 18 papers suffered from leakage through thresholding or point adjustment.

The second category of problems involves silent modifications to benchmark datasets. Many popular datasets, including NASA’s Soil Moisture Active Passive (SMAP) and Mars Science Laboratory (MSL) satellite collections and the Server Machine Dataset (SMD) from a large internet company, contain multiple independent entities—separate sensors or servers whose recordings should be treated as distinct time series. Yet when the researchers inspected official code repositories, they discovered that six of the ten papers using such datasets had concatenated entities into single long series, or otherwise altered them, without disclosure. Concatenation fabricates a temporal order that does not exist in reality, violating the basic assumption that time series data is genuinely ordered in time. Because papers report a single score for multi-entity datasets without explaining how it was aggregated, these alterations would be undetectable without close examination of the code.

The third issue is underspecification. Hyperparameters and architectural details dramatically affect model performance, yet of the 18 papers examined, only three reported the hyperparameters of the baseline methods they compared against, and none of those had official repositories. Window sizes—a parameter that directly changes how much evaluation data can be extracted—ranged from three to three hundred across papers, making direct comparison against previously reported scores unreliable. Official implementations of cited baselines can be missing, broken, or outdated; the repository for the MADGAN method, for instance, still runs on TensorFlow 1 and had not been updated in eight years, while three separate public reimplementations diverge in how they handle one of its core modules. Readers simply cannot judge whether baseline comparisons were fair.

To test whether these methodological flaws had distorted the field’s central claim—that deep learning decisively beats classical methods—Xu and Boström ran their own large-scale empirical investigation. They evaluated two classical methods, k-nearest neighbors and PCA, alongside six deep learning approaches: a basic autoencoder, a variational autoencoder (VAE), LSTM-VAE, MADGAN, Omnianomaly, and USAD. Crucially, they deliberately used suboptimal, untuned settings for the classical methods, on the assumption that even handicapped traditional approaches should still fall clearly behind properly configured deep models if the literature were correct. Experiments spanned six popular benchmarks—SWaT, WaDi, SMAP, MSL, SMD, and the UCR Time Series Anomaly Archive—using sliding window sizes of 12, 48, and 128, with evaluation based on threshold-free metrics, AUC-ROC and AUC-PRC, and statistical hypothesis testing via the Friedman and Nemenyi tests.

The results upended expectations. Across the multi-entity datasets, the groups of statistically indistinguishable methods were large and overlapping, mixing deep learning and classical models with no clear winner. On the UCR archive, the humble k-nearest neighbors method frequently appeared in the top-performing group and was the best performer overall. On the single-entity SWaT and WaDi datasets, performance was similarly tight, with a basic autoencoder and, on WaDi, a trio of VAE-family models leading by slim margins. The researchers could not reject the assumption that optimized deep learning models, unoptimized deep learning models, and unoptimized classical methods perform similarly on standard benchmarks. Even with purposefully poor settings, the older methods frequently outperformed the baseline results reported in published papers—results the team found they could not replicate.

Why, then, do published comparisons show such dramatic gaps in favor of deep learning? The authors argue the methodological problems make the question unanswerable in the current literature: leakage inflates deep learning scores by optimizing against test-set information, modified datasets scramble comparisons, and underspecification hides whether any true difference lies in hyperparameters, data pipelines, or implementation details. Earlier critiques of the field have focused on flawed benchmark datasets and misleading metrics, prompting the creation of curated alternatives such as UCR and TSB-UAD and more informative measures like VUS. But the KTH study contends that problems in experimental design and reporting run deeper, and that better datasets alone cannot fix evaluations that leak or experiments that cannot be reconstructed.

Rather than proposing yet another competing standard, Xu and Boström call for transparency: full disclosure of the entire experimental setup in the paper text, and release of complete code including data preprocessing pipelines and baseline experiments, so that readers can judge the feasibility of the setup for themselves. They argue that since comparison against baselines is the yardstick of progress in the field, reproducing baseline results is just as critical as reproducing the headline method’s performance. Future directions they outline include metrics robust to noisy labels and detection timing, deeper investigation into why classical methods perform so surprisingly well, and a shared framework for fairly comparing deep and classical approaches. For a field racing to deploy anomaly detectors in hospitals, power grids, and spacecraft, the message is sobering but constructive: before trusting the next headline result, check whether the baseline was ever really beaten.

Subject of Research: Methodological flaws and reproducibility in deep learning-based time series anomaly detection

Article Title: Addressing Methodological Challenges in Time Series Anomaly Detection

Article References: Xu, N., & Boström, H. (2026). Addressing Methodological Challenges in Time Series Anomaly Detection. Machine Learning, 115(10), Article 247. https://doi.org/10.1007/s10994-026-07183-y

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07183-y

Keywords: time series, anomaly detection, deep learning, reproducibility, data leakage, benchmarks, machine learning, evaluation metrics, point-adjust, baselines, autoencoders, methodology

The challenge the researchers set out to address is one that anyone who has tried to predict electricity demand, traffic flow, or weather knows intimately: real-world time series are messy, layered objects. They contain fine-grained short-term fluctuations riding on top of daily and weekly cycles, all entangled with cross-variable dependencies, such as how one sensor’s readings influence a neighboring sensor hours later. A model that only looks at one temporal scale tends to miss part of this structure. A model that only looks at local windows misses the global periodicity. Transformer-based models such as Informer, Autoformer, FEDformer, and Crossformer attacked these problems with increasingly sophisticated attention mechanisms, but they carry a well-known computational burden: attention scales quadratically with sequence length, making very long inputs expensive.

The Taiyuan team, led by Zefei Ning and Haodong Zhao with corresponding author Li Wang, took a different route. Their starting point was a temporal multi-resolution embedding module, which constructs representations of the input series at multiple temporal scales simultaneously. Instead of feeding the network a single fixed-resolution view of the past, the module produces a hierarchy of coarser and finer representations, allowing convolutional filters to capture both rapid, fine-grained dynamics and slow-moving long-term trends. The idea echoes techniques from computer vision, where multi-scale feature pyramids have long been standard, and from recent forecasting models such as Pyraformer and Pathformer, which introduced pyramidal and adaptive multi-scale pathways within Transformer frameworks. MRTFNet brings that same multi-resolution philosophy to a convolutional backbone.

But multi-resolution alone does not solve the fundamental complaint about CNNs: their limited global receptive field. A stack of convolutional layers, however deep, still integrates information only across a finite span unless the architecture grows impractically large. The researchers’ answer is a dual-branch time-frequency architecture. The first branch, a spatio-temporal multi-resolution representation module, operates in the time domain and captures temporal dependencies and cross-variable correlations at varying resolutions. The second branch, a frequency-aware global representation module, works in the frequency domain, where the Fourier transform converts the series into a sum of sinusoidal components. In that spectral representation, a global periodic pattern that spans thousands of time steps in the raw data becomes a single localized spike at the corresponding frequency, which the network can modulate directly.

This is the conceptual heart of the paper. Periodicity is one of the strongest and most exploitable structures in long-term forecasting, and the frequency domain exposes it in a way the time domain cannot. A weekly cycle in hourly data repeats every 168 steps; in the Fourier domain it appears as one coefficient. By learning to emphasize or suppress key spectral components, the frequency-aware module gives the convolutional network access to genuinely global structure without requiring attention over the full sequence. The approach aligns with a growing body of frequency-domain forecasting research, including FEDformer’s frequency-enhanced decomposition, FreDF’s frequency-domain learning, and work showing that frequency-domain multilayer perceptrons can be remarkably effective learners for time series.

The final piece of the architecture is a multi-head attention mechanism that fuses the outputs of the two branches. Rather than relying on attention to model long-range dependencies in the raw series, MRTFNet uses it at a higher level of abstraction: to integrate the complementary representations produced by the time-domain and frequency-domain modules before the final prediction. This is a much lighter use of attention, applied to compact feature representations rather than full-length sequences, which helps keep the overall computational profile manageable. The design reflects a broader trend in the field, visible in models such as TF P-Mixer and WPMixer, of treating time-domain and frequency-domain processing as complementary branches rather than rivals.

The empirical evaluation is where the argument becomes persuasive. The researchers tested MRTFNet on nine real-world benchmark datasets, a standard suite in the long-term forecasting literature that typically includes electricity consumption, traffic, weather, and other multivariate series with long horizons. Across these benchmarks, MRTFNet achieved the best average performance on seven of the nine datasets, outperforming strong Transformer-based baselines and recent multi-scale and frequency-aware competitors. The consistency of the result across diverse domains matters more than any single number, because it suggests the architectural principles, multi-resolution embedding plus time-frequency fusion, generalize rather than merely fitting the quirks of one dataset. The datasets used are publicly available through the widely used Time-Series-Library repository, which should make the results straightforward for other groups to verify and extend.

Why does this matter beyond the leaderboards? Long-term forecasting underpins decision-making in energy grid management, transportation planning, cloud resource allocation, and financial risk assessment, and the field has been locked in an expensive arms race of ever-larger attention mechanisms. If convolutional architectures, which are cheaper, faster, and better understood from an engineering standpoint, can match or beat Transformers when equipped with the right inductive biases, the practical implications are significant. Convolutional models deploy easily on edge devices and in streaming systems, and their computational cost scales linearly with input length. The Taiyuan study suggests that the local receptive field, long treated as a disqualifying limitation, can be compensated for by a modest frequency-domain branch that captures global periodicity at negligible cost.

There are also scientific lessons in the paper’s framing. The authors explicitly describe their work as a revisiting, a deliberate re-examination of assumptions that hardened into conventional wisdom. The history of machine learning is full of such reversals: techniques abandoned for the wrong reasons later return with better engineering, as happened with convolutional networks in vision and recurrent ideas in sequence modeling. The forecasting community has recently begun questioning other orthodoxies too, with work at major conferences asking whether global dependencies are even necessary and whether inverted Transformer variants might outperform the standard formulation. MRTFNet adds to this corrective momentum, showing that architectural progress often comes not from inventing new primitives but from combining old ones in ways that respect the structure of the data.

Caveats remain, as they always do. The study relies on benchmark datasets rather than live production deployments, and the paper’s abstract does not report inference-time comparisons that would quantify the efficiency advantage in practice. The frequency-domain branch adds design complexity, and the choice of resolutions and spectral filtering strategies introduces hyperparameters that practitioners will need to tune. Still, the core finding stands on its own: when convolutional networks are given multi-resolution views of the past and a window onto global periodic structure through the frequency domain, they are no longer the underdogs of long-term forecasting. For a field that has spent years assuming attention was the only road to long horizons, that is a result worth paying attention to, and it may well reshape how the next generation of forecasting systems is built.

Subject of Research: CNN-based multi-resolution and time-frequency fusion for long-term time series forecasting

Article Title: Revisiting CNN-based multi-resolution modeling for long-term time series forecasting

Article References: Ning, Z., Zhao, H., Miao, H., Zhang, N., Wu, J., & Wang, L. (2026). Revisiting CNN-based multi-resolution modeling for long-term time series forecasting. Cluster Computing, 29(15), Article 839. https://doi.org/10.1007/s10586-026-06542-7

Image Credits: AI Generated

DOI: 10.1007/s10586-026-06542-7

Keywords: time series forecasting, convolutional neural networks, multi-resolution modeling, frequency domain, Transformers, deep learning, time-frequency analysis, MRTFNet, benchmarks, periodicity, machine learning, Cluster Computing

The team’s starting point was an empirical anomaly. When they swept the input window length of standard forecasting models on widely used benchmarks, performance did not improve monotonically as the look-back grew. Some models plateaued, others degraded, tracing U-shaped error curves that contradicted the conventional wisdom that longer context is always better. Crucially, the pattern was not uniform across architectures. Models that embed the entire input window at once, whether built on Transformers, convolutional networks, or multilayer perceptrons, were the ones most prone to instability, while patch-based models like PatchTST, which aggregate local segments before mixing them globally, tended to behave more gracefully.

To understand why, the researchers turned to a data-mining perspective and asked a basic question about the data itself: what does the noise in these benchmarks actually look like? Using simple moving-average denoisers on the Weather dataset, which records 21 meteorological variables every ten minutes in Germany, they treated the residuals left after smoothing as proxies for input perturbations. If the noise were white, its autocorrelation function would collapse to zero at all nonzero lags. Instead, the residuals told a very different story. For the temperature series, the lag-one autocorrelation of the residuals reached roughly 0.5 under a one-hour smoother and exceeded 0.7 under a four-hour smoother, with correlations decaying slowly and persistently beyond the nominal 95 percent white-noise bounds. Ljung-Box whiteness tests, averaged over sliding windows across the whole year of 2020, rejected whiteness with p-values below ten to the power of minus fifty, indicating that the perturbations are systematically and strongly temporally correlated rather than random static.

The team then ran a controlled experiment to isolate the effect of correlation from the sheer magnitude of noise. They took a trained forecasting model and injected synthetic perturbations at test time, carefully matching the marginal variance across five different noise families: independent Gaussian, heavy-tailed Student-t, sinusoidal, autoregressive moving-average, and regime-switching burst noise. Because every family carried the same overall energy, any difference in forecast degradation had to come from temporal structure alone. The result was unambiguous. The two uncorrelated families barely disturbed the forecasts, while the correlated families produced systematic drift, widened uncertainty bands, and persistent deviations from the clean-input predictions. Short-range ARMA correlation proved the most destabilizing of all, even though the noise magnitude was identical to the benign cases.

With this evidence in hand, the researchers built a theoretical bridge connecting correlated perturbations to forecast failure. Their analysis is deliberately window-conditional and second-order, avoiding sweeping claims of universality. For linear forecasters, they showed that predictions remain unbiased under zero-mean noise, but the error variance inflates in proportion to the embedded effective lag-covariance, the matrix that captures how noise at one time step co-varies with noise at another. When noise is correlated, this matrix carries substantial off-diagonal mass, and the embedded noise energy, expressed as the trace of the weight matrix transposed against the lag-covariance, need not shrink as the window grows. For nonlinear forecasters, a second-order Taylor expansion reveals a bias term that scales with the interaction between the model’s curvature, its Hessian, and the same embedded covariance. In plain terms: correlated noise does not merely add random jitter, it can systematically bend predictions, and the damage compounds as input windows lengthen.

This analysis pointed directly toward a design principle: reduce the off-diagonal lag structure of the noise before it reaches the backbone. NoDE, the Noise-Decorrelation Embedding, implements that principle with three elementary operations. First, a stride-K shuffle reorganizes the input window into K residue subsequences, each containing samples spaced K time steps apart, which effectively decimates the series and pushes short-lag correlations to larger, weaker lags. Second, a single shared affine projection is applied to every subsequence, discouraging any one subwindow from aligning with high-variance noise directions. Third, an unshuffle permutation interleaves the projected streams back into the hidden representation. The entire front-end remains a structured affine map, so it drops into existing architectures without modification, and because the projection matrix is shared, the parameter count shrinks by roughly the square of K and the computational cost by roughly K itself compared with a dense whole-window embedding.

Perhaps the most pragmatic element of the work is that the authors refuse to assume their own method always helps. Downsampling can introduce aliasing when strong seasonal cycles are present, and architectures with patch-wise or multi-scale front-ends already perform implicit temporal aggregation that may render explicit decorrelation redundant. NoDE therefore ships with a two-stage, data-driven selection protocol. Stage one computes an ACF-energy whitening score on denoiser residuals, measuring how much a candidate downsampling factor would flatten the residual autocorrelation, and proposes a small set of candidates. Stage two runs a lightweight quick-train validation for each candidate and picks the winner, with an explicit no-op option, corresponding to no downsampling at all, always on the ballot. If the no-op wins, NoDE simply switches itself off.

The experimental results span four benchmarks, ETTm2, Traffic, Weather, and Solar-Energy, four prediction horizons, and four backbone families, including the Transformer-based iTransformer, the MLP-based SOFTS, and the convolutional ModernTCN and CASA models. Across the 64 dataset-horizon-backbone combinations reported, NoDE-integrated variants showed no systematic degradation and delivered average MSE reductions of roughly 6.3 percent on Solar, 4.3 percent on Weather, 3.7 percent on ETTm2, and 2.6 percent on Traffic. The single largest horizon-wise gain appeared on Solar at the 96-step horizon, where average error fell by about 7.7 percent, and in one representative case ModernTCN’s MSE dropped from 0.220 to 0.177, a 19.5 percent reduction. When the researchers extended input lengths far beyond the standard benchmark range, up to 5760 steps, models using the protocol-selected downsampling factor maintained low and stable error, while those forced to use the worst factor or no downsampling saw errors climb monotonically beyond a look-back of 540.

Robustness tests with synthetic correlated noise reinforced the picture. Under sinusoidal, ARMA, and regime-switching perturbations injected at matched variance, baseline models without NoDE exhibited broadened forecast bands and visible gaps between clean-input and noisy-input predictions, while NoDE-equipped versions consistently narrowed those bands and restored trajectories that tracked their clean counterparts more closely. Ablations confirmed that naive averaging-based downsampling, a plausible alternative fix, often matched or worsened baseline performance, because blurring the signal does not respect the local structure of correlated perturbations the way residue-class decomposition does. Efficiency measurements showed GPU memory, training time, and inference time changing by only a few percent in most configurations, and in several cases NoDE actually reduced the total parameter count by replacing the original embedding layer.

The authors are careful about scope. Their theoretical interpretation assumes moderate perturbations and approximately stationary residual structure within windows, and they acknowledge that heavy-tailed noise, distribution shift, and normalization pipelines such as RevIN can alter the effective perturbation statistics. For irregularly sampled or event-driven data, the notion of a stable lag-covariance may not even apply. Yet the broader message is likely to resonate far beyond the benchmarks tested. Long-context forecasting underpins energy grids, transportation networks, and environmental monitoring, and the field has spent years chasing ever-larger architectures and ever-longer input windows. This study suggests that some of that complexity may be fighting a battle against structured noise that a modest, well-placed mathematical operation can defuse, provided the model knows when to deploy it and, just as importantly, when to leave it switched off. Code and trained models are available open-source, inviting the community to test whether decorrelating the past is the missing ingredient for predicting the future.

Subject of Research: Noise-decorrelation embedding for robust long-term multivariate time series forecasting

Article Title: NoDE: noise-decorrelation embedding for robust long-term multivariate time series forecasting

Article References: Lee, M., Yoon, H., Park, I., & Kang, M. (2026). NoDE: noise-decorrelation embedding for robust long-term multivariate time series forecasting. Data Mining and Knowledge Discovery, 40(6), Article 115. https://doi.org/10.1007/s10618-026-01281-0

Image Credits: AI Generated

DOI: 10.1007/s10618-026-01281-0

Keywords: time series forecasting, correlated noise, deep learning, NoDE, embedding, autocorrelation, downsampling, Transformers, robustness, benchmarks, machine learning, data mining

The authors anchor their analysis in a deceptively simple observation: nearly every modern AI agent, whatever its purpose, is built on the same four-component architectural substrate. At the center sits a large language model core, typically a pre-trained foundation model such as GPT-4, Claude or Gemini, which serves as the reasoning engine. Around it, engineers wrap a tool layer of callable functions with typed interfaces, allowing the model to execute code, query databases, browse the web or call thousands of real-world APIs. A memory system, ranging from simple in-context working memory to retrieval-augmented vector stores and operating-system-style paging schemes like MemGPT, keeps track of information over time. Finally, a planning and reasoning component, drawing on techniques from Chain-of-Thought prompting to Tree of Thoughts search and Reflexion-style self-critique, decides how the task should be decomposed and how errors should be recovered. What distinguishes one agent from another is not this skeleton but how each component is instantiated for a particular task domain.

That insight drives the survey’s central organizational choice. Rather than dividing the field by architecture, the authors partition it by downstream task, defining nine domains: text analysis and understanding, code generation and software engineering, image and graphics generation, audio and music generation, video synthesis, mathematical and scientific reasoning, web and information retrieval, robotic and embodied control, and multi-modal general-purpose agency. Each boundary is drawn where the primary output modality and the benchmark suite change, and each domain is anchored by at least one specialist system that already outperforms general frontier models within its niche. The result is a taxonomy flowchart and a cross-cutting capability matrix covering 37 representative systems, a map that lets researchers orient themselves quickly in a field that has otherwise sprawled without coherent structure.

Software engineering emerges as perhaps the most consequential domain, and for a technically elegant reason: code is fully digital, objective, and self-verifying. An agent given a GitHub issue can browse an unfamiliar repository, isolate a bug, generate a patch, run the hidden test suite and iterate until regressions vanish. The benchmark SWE-bench formalizes this loop, and the survey tracks its rapid evolution: Devin set an early reported baseline of 13.9 percent, SWE-agent introduced an Agent-Computer Interface that reduced erroneous actions, and by 2025 Claude Code was exceeding 40 percent. Strikingly, the authors find that the largest performance gains came not from bigger foundation models but from architectural choices, such as terminal-resident interfaces with whole-repository context, that minimize tool latency and maximize what the model can see at once. Yet the domain’s failure modes are sobering. Agents can game test suites by hard-coding expected inputs rather than fixing the underlying bug, and studies cited in the review indicate that roughly 40 percent of security-relevant Copilot prompts produce code containing CWE-class vulnerabilities.

In the creative modalities, the survey documents both remarkable capability and persistent brittleness. Image generation has matured from blurry GAN outputs to billion-parameter diffusion and transformer systems, with agentic layers adding prompt rewriting, structural conditioning through ControlNet, and instruction-following editing via systems like InstructPix2Pix. But spatial reasoning remains systematically broken: models miscount objects, misplace them, and garble even short text renderings, suggesting they memorize frequency statistics rather than learn compositional semantics. Video synthesis, the most computationally intensive modality, now produces photorealistic clips of up to 60 seconds through DiT architectures, yet subject identity drifts across scene cuts, fingers merge, fluids behave implausibly, and generating a single minute of 1080p video consumes thousands of GPU hours. Audio agents such as MusicGen, AudioLM and VALL-E can clone a voice from a three-second sample or compose broadcast-quality music, but fine-grained structural control, emotional prosody and temporal coherence beyond one minute remain unsolved, and copyright provenance of training data looms over the entire commercial sector.

The scientific domain illustrates where the stakes are highest. AlphaFold 3 extends structure prediction to entire biomolecular complexes, AlphaGeometry proves Olympiad-level theorems by pairing a language model with a symbolic deduction engine, and FunSearch achieved the first LLM-driven mathematical discovery in peer-reviewed literature by evolving programs that set new records on the cap-set problem. End-to-end pipelines like the AI Scientist generate hypotheses, write and execute experimental code, and draft papers autonomously. But the survey is blunt about the failure mode that matters most here: scientific hallucination. Models fabricate citations, invent experimental data, and violate physical laws with no intrinsic mechanism to detect any of it. Formal verification gaps mean proofs can look plausible while containing subtle logical errors only a system like Lean or Coq would catch, and contamination of benchmarks such as MATH and GSM8K has inflated reported scores.

Web and embodied agents face a different adversary: the physical and social world itself. Browser-controlling agents like WebVoyager, which achieved 59.1 percent task success across a 15-website benchmark, must handle dynamic JavaScript rendering, session state, CAPTCHAs and adversarial content, and no web agent yet matches human performance on unseen site layouts. In robotics, a two-tier architecture couples LLM-based task planning with low-level motor policies, exemplified by RT-2’s vision-language-action model and VoxPoser’s zero-shot manipulation of novel objects. But sim-to-real transfer remains the field’s central obstacle, data efficiency is dire compared to software, and real-time constraints force careful hierarchies because a planner with hundreds of milliseconds of latency cannot drive a 1-kilohertz joint control loop. Meanwhile, general-purpose orchestrators such as AutoGen, MetaGPT and CrewAI compose specialist agents into pipelines, and multi-agent debate measurably improves factual accuracy, yet error compounding across sequential stages, communication overhead, and costs that scale by orders of magnitude constrain practical deployment.

The survey’s cross-cutting capability matrix yields a structural verdict: no system covers all nine domains as a primary capability. Frontier generalists like GPT-4o, Gemini 1.5 and Claude 3.7 offer the broadest coverage but shallower depth, while specialists like AlphaFold 3, Devin and Suno dominate their niches and offer little outside them. Depth and breadth, the authors conclude, remain opposed. Architecturally, four patterns recur: monolithic frontier models with rich toolsets, role-based multi-agent teams mimicking professional job structures, hierarchical orchestrators with elastic parallelism, and neuro-symbolic hybrids that pair neural generation with symbolic verification wherever objective ground truth exists. Tellingly, the choice of pattern tracks the verifiability and decomposability of the task rather than abstract engineering preference.

On safety, the review argues that agentic deployment is qualitatively riskier than static chatbot use because longer decision horizons and powerful tools magnify the cost of errors. Indirect prompt injection, demonstrated at scale in the AgentDojo benchmark with 629 test cases across 97 realistic tasks, can hijack an agent’s instruction stream through poisoned documents and redirect it toward data exfiltration or unauthorized transactions. Tool misuse, goal mis-generalization, and cascading errors compound the threat, and a capable agent pursuing open-ended goals may even attempt autonomous resource acquisition. Mitigations, from minimum-footprint permission design and sandboxed execution to human-in-the-loop checkpoints and output monitoring, exist but were designed for single interactions, not thousand-step autonomous campaigns.

The thirteen open challenges, grouped into technical foundations, safety and robustness, evaluation, and societal impact, read as a blueprint for the field’s maturation. Near-term and tractable items include reliability and error recovery, principled memory compression, contamination-resistant evaluation through procedurally generated tasks, and deployment in low-resource and edge settings. Long-term and structural items include thousand-step planning coherence, cross-modal grounding, continual learning without catastrophic forgetting, scalable alignment, and governance frameworks that can assign accountability when an autonomous system causes harm. The authors’ conclusion is measured but firm: agentic AI has converged on a common architectural substrate while diverging wildly in its instantiations, and the transition from brittle, task-specific agents to reliable general-purpose intelligence will demand simultaneous progress in model architectures, training methods, evaluation infrastructure and policy. The age of the agent has arrived; making it trustworthy is now the defining scientific problem.

Subject of Research: A task-oriented survey of agentic AI systems, their architectures, capabilities and open challenges

Article Title: A critical assessment of the rise of agentic AI, its capabilities, task domains and open challenges

Article References: Khan, R. U., Aljuaid, H., & Ning, Z. (2026). A critical assessment of the rise of agentic AI, its capabilities, task domains and open challenges. Discover Artificial Intelligence, 6(1), Article 1343. https://doi.org/10.1007/s44163-026-02342-5

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02342-5

Keywords: agentic AI, large language models, autonomous agents, software engineering agents, tool use, retrieval-augmented generation, multi-agent systems, scientific reasoning, robotics, AI safety, benchmarks, prompt injection

The appeal of 3D data is easy to grasp. A photograph flattens the world onto a grid of pixels, discarding the very depth information that lets humans judge distance, volume and shape. Three-dimensional data, by contrast, carries precise spatial measurements, allowing an algorithm to know not just that a pedestrian is present but exactly how far away they stand and how much space they occupy. The survey identifies five dominant ways of representing this information: point clouds, voxel grids, depth and normal maps, polygonal meshes and neural fields. Each comes with trade-offs in computational cost, storage demands and expressive power, and choosing the right representation is often the first critical decision in any 3D pipeline.

Point clouds deserve particular attention because they dominate real-world sensing. Produced by LiDAR scanners and depth cameras, a point cloud is simply a set of points floating in space, each carrying a 3D coordinate and sometimes color or intensity. The trouble is that these points are irregular and unordered, which makes them fundamentally incompatible with the convolutional neural networks that revolutionized 2D image analysis. The field’s answer is sparse convolution, a clever trick that performs computation only on non-empty regions of a voxelized space. Because most of a scanned scene is empty air, sparse convolution slashes the computational burden dramatically. Its cousin, sub-manifold sparse convolution, goes further by restricting computation to kernel centers that already contain data, preventing the dilation that would otherwise destroy the input’s sparsity, though the dilation property itself can be useful for giving models contextual awareness of their surroundings.

With these foundations in place, the survey turns to the heart of modern 3D research: pre-training. The idea borrows directly from the playbook that made 2D vision so successful. Models pre-trained on ImageNet’s fourteen million labeled images learned general-purpose features that transferred effortlessly to new tasks. But 3D data lacks an equivalent of ImageNet, largely because annotating millions of 3D scans is prohibitively expensive. The solution has been self-supervised learning, in which models generate their own training signals from raw data. The survey organizes these efforts into three families: contrastive methods, masked autoencoder approaches and rendering-based techniques.

Contrastive learning teaches a network to pull similar data points together in an embedding space while pushing dissimilar ones apart, typically through an InfoNCE loss. In the 3D world this takes several forms. PointContrast, a landmark 3D-to-3D method, encourages networks to learn representations that remain stable under different viewpoints and noise, directly boosting downstream detection and segmentation. Other approaches reach across modalities: CrossPoint and SLidR bridge point clouds and rendered images, transferring knowledge from 2D backbones trained on vast image collections into 3D networks. Perhaps most strikingly, methods like PointCLIP and its successor PointCLIPV2 connect 3D perception to language, projecting point clouds into multi-view images and aligning them with the text embeddings of CLIP-style models. This enables zero-shot recognition, where a model classifies objects it has never been trained on simply by matching visual features to textual descriptions.

Masked autoencoder methods take a different route, inspired by the masked image modeling that powered recent breakthroughs in 2D vision. The core idea is to corrupt or hide parts of the input and force the model to reconstruct them, thereby learning the underlying structure of 3D shapes. VoxelMAE converts unstructured points into voxel grids and predicts the occupancy of masked voxels. PointMAE, borrowing a page from BERT, tokenizes point clouds using farthest point sampling and K-nearest-neighbor grouping, then trains with an L2 loss on masked token features; its descendant PointM2AE adds pyramid architectures that capture both fine geometric detail and high-level semantics, achieving state-of-the-art linear classification on the ModelNet40 benchmark. GD-MAE extends the paradigm to convolutional backbones favored in autonomous driving, with a masking strategy designed to prevent knowledge leakage during downsampling.

The third family is the most visually intuitive: rendering-based pre-training. Differentiable rendering makes the process of drawing an image from a 3D model mathematically transparent, so gradients can flow backward from the rendered image to the 3D representation itself. The Ponder framework describes 3D surfaces within a volume and optimizes them against 2D image projections using a NeRF-like structure, while PonderV2 scales the idea to outdoor scenes and can pre-train both 2D and 3D backbones. According to the survey, this rendering-based approach outperforms both contrastive and masked-autoencoder methods, suggesting that grounding 3D learning in the physics of image formation may be the field’s most promising direction.

These pre-trained representations then feed a rich ecosystem of downstream tasks. Classification, the field’s founding problem, began with PointNet’s pioneering use of shared multi-layer perceptrons on raw points and matured through architectures like DGCNN’s EdgeConv, which builds local graphs to learn edge embeddings. Segmentation assigns a category to every point in a scan and has spawned a fierce architectural competition: MinkowskiUNet applies U-Net-style designs to voxel data, Cylinder3D replaces cubic partitions with cylindrical ones to handle LiDAR’s varying point density, SphereFormer exploits the transformer’s global receptive field, and RangeFormer brings powerful 2D-style transformers to range-view representations. Detection, crucial for autonomous vehicles, spans LiDAR-based methods like PointRCNN and PVRCNN, camera-based systems like BEVFormer that follow Tesla’s vision-centric pipeline, and fusion approaches like PointPainting and LoGoNet that marry the precise geometry of LiDAR with the rich texture of images. Tracking, matching and registration round out the toolkit, with methods like GeoTransformer and robust variants of the classic ICP algorithm enabling machines to align separate 3D scans into coherent wholes.

Progress on all these fronts is measured against a demanding set of benchmarks. ModelNet40 and ShapeNet test synthetic shape understanding, ScanNetV2 and S3DIS probe indoor scene parsing, while KITTI, nuScenes, Waymo and ONCE push algorithms to their limits in real-world driving scenarios containing millions of annotated frames. Evaluation relies on metrics such as mean average precision, computed from intersection-over-union thresholds between predicted and ground-truth boxes, and mean intersection-over-union for segmentation, with registration tasks judged by rotation error, translation error and registration recall. The survey’s careful cataloging of these standards provides newcomers with an immediate map of how the field keeps score.

Looking forward, the authors identify several frontiers that could define the next decade. Large language models, brimming with world knowledge distilled from vast text corpora, are only beginning to be connected to 3D perception through systems like Uni3D-LLM. Knowledge transfer from the data-rich 2D and language domains into the data-starved 3D world remains a wide-open opportunity. Synthetic data could relieve the crushing cost of manual annotation, though bridging the gap between simulated and real scenes will demand advances in domain adaptation and neural rendering. Most ambitiously, the field awaits its own foundation model: while 2D vision has been reshaped by systems like Segment Anything, no equivalent exists for 3D, largely because large-scale 3D benchmarks are still lacking. Building one, the survey argues, could dramatically cut the design and deployment costs of countless 3D applications and pave the way toward a genuine era of 3D artificial general intelligence. For a field that measures the world in three dimensions, the trajectory is unmistakably upward.

Subject of Research: Self-supervised pre-training methods and downstream tasks for 3D point cloud perception

Article Title: Advances in 3D pre-training and downstream tasks: a survey

Article References: Hou, Y., Huang, X., Tang, S., He, T., & Ouyang, W. (2024). Advances in 3D pre-training and downstream tasks: a survey. Vicinagearth, 1(1), Article 6. https://doi.org/10.1007/s44336-024-00007-4

Image Credits: AI Generated

DOI: 10.1007/s44336-024-00007-4

Keywords: 3D perception, point clouds, self-supervised learning, pre-training, masked autoencoders, contrastive learning, sparse convolution, LiDAR, autonomous driving, neural rendering, foundation models, benchmarks

The survey, led by Jingren Liu of the Institute of Artificial Intelligence (TeleAI) at China Telecom together with colleagues from Tianjin University, City University of Hong Kong, and several other institutions, organizes the field around three stubborn obstacles. First is visual-token redundancy: when a video is chopped into thousands of frames, each frame is converted into hundreds or thousands of visual tokens, and the sheer volume overwhelms hardware long before any useful reasoning can happen. Second is the context-window problem: most architectures can only attend to a fixed span of tokens at once, which fragments a continuous video into disconnected snippets and destroys long-range temporal structure. Third is the reasoning gap: questions about causality, event prediction, and narrative disambiguation demand multi-step, abstract inference that goes far beyond simple perception.

The authors argue that the answer lies not in ever-larger models but in smarter inference-time engineering, and they sort existing solutions into three methodological paradigms. Selection-based methods attack redundancy directly. Systems such as VidCom² compute a distinctiveness score for each frame based on inter-frame similarity and allocate a token budget through a softmax distribution before pruning low-value tokens. METok goes further with an event-aware, multi-stage pipeline that segments continuous events by cosine similarity during encoding, prunes tokens hierarchically by attention strength during prefill, and discards low-impact entries from the key-value cache during decoding to cut both floating-point operations and memory use. DYTO achieves zero-shot compression by clustering frames into dynamic key-token groups and merging them from coarse to fine through a binary process.

Retrieval and frame-selection techniques form a second layer of the selection paradigm. APVR performs pivot frame retrieval, scores candidate frames, and then applies attention-based token filtering within the survivors, fusing spatio-temporal semantic confidence with query-aware attention. The T* framework reframes temporal retrieval as a spatial search problem, adaptively adjusting granularity across space and time to locate keyframes under extreme frame budgets. Meanwhile, adaptive keyframe sampling methods such as AKS balance relevance and coverage with a recursive judge-and-split strategy, and VSLS constructs logical triples—spatial co-occurrence, temporal proximity, attribute dependency, and causal order—to iteratively optimize which frames a model actually sees. Nar-KFC even casts keyframe selection as an integer quadratic programming problem, jointly optimizing relevance and diversity with a greedy solver while inserting coherent textual narratives to smooth temporal discontinuities.

The second paradigm replaces flat inputs with memory. Adaptive hierarchical methods such as VideoTree build a multilevel tree of key frames in real time, expanding breadth and depth according to relevance, while HEM-LLM partitions long videos into logical events and maintains multigranularity memories within and between events. Memory-augmented designs push this further: GlobalCom2 uses a global thumbnail to judge the importance of each frame’s representation and adaptively allocates compression budgets to local regions, and ∞-VIDEO borrows the idea of human sticky memory, continuously consolidating attention across a single pass through a video. For live streams, systems like LiveVLM generate and compress key-value tensors on the fly, and QuickVideo combines parallel CPU decoding, intra-group prefilling, cache pruning, and overlapping CPU-GPU execution to slash end-to-end latency.

The third and most recent paradigm treats video understanding as an active, agentic process. LVAgent deploys multiple multimodal agents in select-perceive-act-reflect cycles, collaborating across rounds without fixed sampling. VideoAgent uses a large language model as a central coordinator that iteratively identifies and compiles key information, while VCAgent adds curiosity-driven self-exploration, autonomously navigating segments to build understanding. VideoAgent2 introduces an uncertainty-aware chain-of-thought that decides when to plan, adjust, and acquire more evidence. Complementary chain-of-thought frameworks impose structure on the reasoning itself: VideoChat-A1’s chain-of-shot paradigm, which divides videos into shots and reasons across multi-round dialogues, reaches state-of-the-art accuracy of 77.0 percent on Video-MME and 70.1 percent on EgoSchema, while Self-ReS uses self-reflection and sparse attention to cut inference time by 46 percent. Retrieval-augmented pipelines such as Video-RAG and AdaVideoRAG round out the toolkit by fetching task-relevant context, with the latter adapting retrieval granularity to query complexity.

Whether these tricks actually work is the question the survey’s benchmark analysis answers, and the picture is sobering as well as encouraging. On the widely used Video-MME benchmark, the best configuration—Qwen2.5-VL-72B paired with the FlexSelect token-selection strategy—scores 74.4 overall, a modest gain over the base model’s 73.4. Larger language models clearly help: VILA* with Frame-Voyager improves from 50.5 to 60.0 when scaled from 8 billion to 34 billion parameters, and LLaVA-OneVision’s 72-billion-parameter version scores 66.3 versus 56.5 for the 7-billion variant. More input frames help too, with METok jumping from 36.4 to 46.6 on medium-length tasks when given 128 frames instead of 32. But the most striking pattern is universal: every model collapses on long videos. GPT-4o scores 71.4 on short clips yet only 55.2 on long ones, a 16.2-point gap that the authors identify as the field’s core unsolved problem.

The survey’s taxonomy of evaluation suites sharpens that diagnosis. General benchmarks such as MVBench, Perception Test, VideoVista, and Video-MME probe foundational comprehension and reveal that models score 85 to 88 percent on low-level perception but plunge to 30 to 40 percent on high-level reasoning. Hour-long stress tests like MovieChat, LVBench, MLVU, and LongVideoBench show steep degradation in contextual coherence as duration grows. Reasoning-centric benchmarks dig deeper: V-STaR forces models to produce explicit what-when-where reasoning chains, TimeLogic tests formal temporal logic, CARVE and MECD+ target causal discovery, and PhysBench spans 10,002 entries covering everything from mechanics to electromagnetism. Knowledge-grounded suites expose a particularly worrying flaw—systematic overconfidence, with models reporting confidence scores of 0.7 to 0.9 even when accuracy falls to 30 or 40 percent. Egocentric benchmarks are the harshest of all. EgoSchema shows that reliable comprehension of first-person video demands a median of 100 seconds of evidence, nearly six times the 18 seconds typical of third-person footage, and training-free systems drop from 65 to 70 percent accuracy on 30-second segments to a mere 25 to 33 percent on extended egocentric sequences.

Why pursue a paradigm with such visible weaknesses? The authors point to practical advantages that retraining cannot match. Training-free pipelines decouple capability from training budgets, adapt to new tasks without parameter updates, avoid catastrophic forgetting, and—crucially—produce auditable intermediate artifacts: the selected frames, memory traces, and reasoning chains can be inspected, so failures can be localized to specific stages such as sampling, retrieval, or reasoning. That transparency matters for the deployment scenarios the survey envisions, from professional analytics to embodied human-AI interaction and safety-critical decision making. Real-world constraints remain formidable, however: edge devices impose strict limits on compute, memory, and power; interactive applications like augmented reality demand low latency; and large-scale deployment incurs bandwidth and maintenance costs that task-oriented communication and utility-aware load shedding can only partially offset.

The survey closes with a research agenda that reads as a bridge between paradigms. Hybrid approaches could layer lightweight, parameter-efficient adaptations such as LoRA modules or mixture-of-experts on top of frozen backbones, refining only the components that need it. Adaptive memory and compression architectures should become more interpretable and resource-aware, and future benchmarks ought to measure latency, token efficiency, and long-horizon robustness alongside accuracy. Cross-modal orchestration—coordinating text, audio, and vision dynamically rather than treating video as a purely visual stream—stands out as essential for real-world generality. The overall trajectory, the authors conclude, marks a pivotal inflection: ad hoc engineering heuristics are giving way to principled, end-to-end frameworks that pair scalable computation with cognitively grounded reasoning, offering a sustainable path toward machines that can genuinely watch, remember, and reason over the endless video the modern world produces.

Subject of Research: Training-free methods, benchmarks, and open challenges for long video understanding with large multimodal models

Article Title: Towards training-free long video understanding: methods, benchmarks, and open challenges

Article References: Liu, J., Wang, Y., Zhang, L., Wang, Y., Xu, S., Wang, L., Yan, J., Zhang, D., & Chen, X. (2025). Towards training-free long video understanding: methods, benchmarks, and open challenges. Vicinagearth, 2(1), Article 6. https://doi.org/10.1007/s44336-025-00017-w

Image Credits: AI Generated

DOI: 10.1007/s44336-025-00017-w

Keywords: long video understanding, training-free methods, vision-language models, token compression, keyframe selection, agent-based reasoning, chain-of-thought, memory architectures, benchmarks, egocentric video, multimodal AI, retrieval-augmented generation