Tag: Mixture of Experts

A study of nearly ten million trips demonstrates that integrating driver learning curves with telematics data reduces auto insurance premium ...

Researchers developed AirMixer, a channel-independent neural network that improves long-term air quality forecasting across eleven benchmark datasets.

A benchmark across 6,900 outputs revealed that reasoning modes substantially increased clinical text extraction accuracy while noticeably lengthening inference latency.

A comprehensive critical review synthesizes the mathematical principles, architectural innovations, and empirical scaling behaviors governing modern Transformer foundation models.

A segment-conditioned Flexible-K Mixture of Experts model lets expert participation vary across timesteps and achieves low RMSE on the C-MAPSS ...

A backpropagation-free Deep Memory mechanism preserves representations in dissipative AI architectures by recording centroids, seeding replaced units and stabilizing through ...

Helios combines congestion-aware expert placement with runtime replication, reporting simulated gains of up to 45 percent in goodput and 52.6 ...

A new multimodal transformer framework detects REM sleep behavior disorder with an AUC of 0.95 while running in 28 milliseconds, ...

A computational study found that confidence-based routing between subject-specific and pooled decoders improved P300 EEG decoding in two public datasets.

A seven-degree-of-freedom robotic arm completed all 30 trials with familiar objects and 52.9% of trials with seven unseen household objects ...