The study, authored by Mengwen Xiao, Kangrui Du and Ziqi Lin, sits at the intersection of two research traditions that have matured rapidly over the past decade. The first is representational alignment, a family of techniques that quantify the similarity between the internal states of artificial neural networks and the activity patterns recorded from living brains. The second is chain-of-thought reasoning, the strategy by which large language models decompose complex problems into intermediate steps before committing to an answer. The new research connects these traditions in a striking way: instead of using brain activity only as a post-hoc yardstick, the authors use it as a training signal, encouraging a model’s internal representations to track those observed in human neural data while it reasons.

To appreciate why this matters, it helps to recall how the relationship between brains and machines has typically been studied. In a landmark 2023 analysis, researchers including Alex Zador and colleagues argued in Nature Communications that understanding natural intelligence requires studying both brains and their artificial analogues together, a perspective that helped galvanize the NeuroAI community. Subsequent work by Greta Tuckute and collaborators, published in Nature Human Behaviour in 2024, showed that language models vary widely in how well their internal states predict neural responses to sentences, and that this predictive power relates to how ‘brain-like’ a model’s representations are. Until recently, however, such alignment scores were largely descriptive. They told researchers which models resembled the brain, but they did not tell the models how to become more brain-like, nor whether doing so would improve them.

The conceptual shift from passive alignment to active neural guidance is the heart of the new work. In the passive paradigm, a trained model is frozen and its hidden activations are compared, layer by layer, with recordings such as functional MRI or magnetoencephalography taken while people read or solve problems. In the active paradigm described by Xiao and colleagues, the comparison happens during learning. The model receives an additional objective that rewards representational similarity to human neural data, effectively telling the network, at each stage of processing, that its internal geometry should not drift too far from the geometry observed in the brain. This turns the brain from a mirror into a compass, and the consequences for reasoning performance are substantial.

The technical machinery behind this idea draws on a rich literature. Representational similarity analysis and related encoding models provide differentiable measures of alignment, meaning that the gap between a model’s internal states and recorded brain activity can be computed as a quantity that gradient descent can minimize. Changde Du and Huiguang He, writing in an accompanying News and Views article in Nature Machine Intelligence, place the new study in the context of their own earlier work, published in the same journal in 2025, which explored how neural signals can inform machine learning systems. The new contribution extends this program from perception and language understanding into the domain of multi-step reasoning, where errors compound across steps and small representational deviations can cascade into dramatically wrong conclusions.

Reasoning is precisely where large language models remain most fragile. Chain-of-thought prompting, introduced by Jason Wei and colleagues at NeurIPS in 2022, showed that asking models to generate intermediate reasoning steps can unlock dramatic gains on arithmetic, logic and symbolic tasks. Yet subsequent analyses, including a large-scale study by Abulhair Saparov and colleagues in 2023, revealed that models often succeed on familiar reasoning patterns while failing catastrophically on novel ones, suggesting that their apparent reasoning may rest on shallow pattern matching rather than robust inference. If human neural representations encode something about the structure of valid inference that current training objectives fail to capture, then aligning with those representations during training could inject exactly the inductive bias that chain-of-thought systems lack.

The evidence reviewed by Du and He suggests that this is more than speculation. When the neural-guided objective is incorporated into training, the resulting models not only show stronger correspondence with human brain activity but also demonstrate improved reliability on reasoning benchmarks, producing more consistent and accurate chains of intermediate steps. The authors of the News and Views piece frame this as a move ‘from passive alignment to active neural guidance,’ a phrase that captures the dual payoff of the approach. Alignment ceases to be a descriptive statistic and becomes a functional ingredient of learning, and the improvement in reasoning suggests that the brain’s representational structure carries information that is genuinely useful for artificial inference, not merely an incidental byproduct of shared training data.

The finding also speaks to a long-running debate about what language models actually learn. Ev Fedorenko, Steven Piantadosi and Edward Gibson argued in Nature in 2024 that formal linguistic competence, the ability to produce fluent grammatical language, can be dissociated from functional linguistic competence, the ability to use language for thinking. Large models have largely conquered the former while remaining uneven at the latter. Neural guidance offers a possible bridge: if the internal representations that support human functional competence can be transferred, even partially, into artificial systems, then machines might acquire not just the form of reasoning but something closer to its substance. Related work by Juan Prado, Annapurni Chadha and James Booth, published in the Journal of Cognitive Neuroscience in 2011, mapped the neural substrates of arithmetic processing in children, illustrating the kind of structured neural data that such alignment approaches can exploit.

Independent work reinforces the momentum behind this direction. A 2026 study by Jie Chen, Yue Qi, Yu Wang and Guang Pan in Nature Communications reported further evidence that neural signals can inform and improve machine learning systems, indicating that the strategy is not an isolated result but part of an emerging research program. Together, these studies suggest that the coming years will see increasingly tight coupling between neuroimaging pipelines and model training loops, with brain data flowing not only into evaluation suites but into the loss functions themselves. The practical challenges remain considerable: collecting high-quality neural recordings at scale is expensive, individual differences in brain organization complicate generalization, and the causal question of why neural alignment helps reasoning is only beginning to be answered.

Still, the implications are hard to overstate. If the human brain can serve as a reliable guide for training artificial reasoners, then neuroscience and machine learning are no longer parallel enterprises exchanging occasional insights but a single, integrated discipline in which each field sharpens the other. Models guided by neural signals could become more trustworthy partners in science, medicine and education, while the act of aligning machines to minds may itself reveal which aspects of neural computation are essential to intelligence and which are biological accident. The work of Xiao, Du and Lin, and the perspective offered by Du and He, mark an early but compelling step in that direction, transforming the humble brain scan from a passive record of thought into an active instrument for teaching machines how to think.

Subject of Research: Using human brain signals to actively guide and improve the reasoning of artificial intelligence models

Article Title: Steering machine reasoning with brain signals

Article References: Du, C., & He, H. (2026). Steering machine reasoning with brain signals. Nature Machine Intelligence. https://doi.org/10.1038/s42256-026-01302-z

Image Credits: AI Generated

DOI: 10.1038/s42256-026-01302-z

Keywords: NeuroAI, representational alignment, large language models, chain-of-thought reasoning, brain signals, machine learning, neural encoding, Nature Machine Intelligence, cognitive neuroscience, artificial reasoning, fMRI, brain-inspired AI

The work follows the idea that neural planning does not rely on brute-force computation to reach a solution. Instead, it uses principles observed in the brain—particularly mechanisms linked to the hippocampus—to generate candidate futures and refine actions toward a goal. “The brain works in a completely different way to today’s AI systems,” explains Wolfgang Maass from TU Graz, emphasizing an algorithmic translation of biological computation.

At the core of the approach are three interacting mechanisms. First, the model forms cognitive maps, converting relationships among abstract entities into geometric structure within neural codes—effectively providing a “sense of direction.” Second, it uses stochastic neural computations that continuously propose hypothetical scenarios, enabling the system to explore without calculating every pathway. Third, it applies compositional coding, breaking down information and action plans into reusable components.

Together, these mechanisms allow the model to “imagine” possible routes and test them in imagination before committing. Rather than searching exhaustively, the system samples intermediate steps: when a randomly selected move appears to align with the goal, guided by the cognitive map, that direction is pursued. At each new position, the model again evaluates options, gradually steering toward the target while maintaining flexibility.

A key claim is adaptability. Because planning is driven by sampling from cognitive maps and compositional representations, the system can respond to changed or newly introduced situations without requiring retraining. This capability contrasts with many conventional pipelines that must be retrained to handle new environmental structures or objectives.

To demonstrate the idea, the team evaluated the model on three challenges: navigating a two-dimensional space, orienting within an abstract multi-dimensional space, and assembling or disassembling a silhouette built from modular blocks. Across tasks, the system showed goal-directed planning behavior consistent with a sampling-and-map framework.

The researchers stress that the approach is not meant to replace today’s large language models. Instead, it proposes an alternative route for applications where efficient local decision-making matters. With further development, brain-inspired planning systems could broaden AI beyond cloud-scale compute.

Such energy-aware methods could be especially valuable for robots, autonomous vehicles, and edge devices—settings where hardware constraints demand low power operation. The study was conducted with collaborations including Tsinghua University and Italy’s National Research Council.

Subject of Research: Brain-inspired neural planning and cognitive-map-based sampling
Article Title: Neural sampling from cognitive maps enables goal-directed imagination and planning
News Publication Date: 21-Jul-2026
Web References: http://dx.doi.org/10.1038/s42256-026-01254-4
References: Nature Machine Intelligence (DOI: 10.1038/s42256-026-01254-4)
Image Credits: Lunghammer – TU Graz

Keywords

Brain-inspired AI; cognitive maps; neural sampling; stochastic planning; compositional coding; energy-efficient machine intelligence

Unlike the cerebrum, which undertakes intensive “thought” processing, the cerebellum specializes in swift reflexes by selectively responding to surprising stimuli. Taking inspiration from this, the researchers designed a memtransistor device capable of operating in two distinct modes: excitatory and inhibitory. This dual functionality mirrors the balance of neural signals in the cerebellum, where excitation and inhibition maintain equilibrium during normal activity and shift rapidly upon detecting novelty.

At the heart of the device’s innovation lies the use of molybdenum disulfide, an atomically thin semiconductor renowned for exceptional electrical properties. The engineers implemented an asymmetric transistor architecture where one electrode slightly overlaps the semiconductor through a thin insulating layer. This structural nuance allows the direction of applied voltage to switch the memtransistor between excitatory and inhibitory responses, effectively emulating synaptic behavior in hardware.

The implications for AI systems are profound. In testing, the device processed electrocardiogram (ECG) data streams, accurately discerning abnormal heart rhythms within milliseconds—faster than twice the speed of current AI methods. By concentrating computational effort solely on atypical inputs instead of continuous data streams, the memtransistor reduces requisite computer operations by approximately 10,000 times, paving the way for ultra-low-power “always-on” AI applications.

Such efficiency gains could revolutionize wearable health monitors by enabling near-instant cardiac anomaly detection, bolster autonomous vehicles’ responsiveness to sudden environmental changes, enhance robotic interaction safety, and tighten cybersecurity systems by catching suspicious activities before escalation—all with minimal energy footprints.

This research advances a broader vision to reimagine AI hardware by collapsing memory and computation into single devices, a principle previously demonstrated by the team using memtransistors for classification tasks at a 100-fold energy reduction. Moving forward, the team aims to incorporate adaptive learning mechanisms to mimic the cerebellum’s capacity to habituate to repeated stimuli, further refining its neuromorphic prowess.

By harnessing atomically thin materials and innovative transistor design, this cerebellum-inspired memtransistor heralds a new era of hardware-efficient novelty detection, embodying a paradigm shift in neuromorphic engineering and energy-conscious AI.

Subject of Research: Cerebellum-inspired memtransistor devices for energy-efficient novelty detection
Article Title: Cerebellum-inspired memtransistors enable emergent differentiation for hardware-efficient novelty detection
News Publication Date: 10-Jul-2026
Web References: http://dx.doi.org/10.1038/s41467-026-75212-4
Image Credits: Mark C. Hersam/Northwestern University

Keywords

Artificial intelligence, Neuromorphic computing, Memtransistors, Cerebellum, Novelty detection, Molybdenum disulfide, Low-power AI, Wearable health monitors, Autonomous robotics, Cybersecurity