ocean concept bottleneck model explains heat predictions
a new model uses physics-based concepts to make ocean heat forecasts interpretable, revealing the drivers behind predictions.
topic
research
learning when to act for safe, communication-efficient rl
a new method learns both control and when to communicate, using a safety shield to reduce sampling while maintaining stability.
quantum algorithm cracks impossible materials problem in seconds
a quantum-inspired algorithm simulates complex quasicrystals with over 268 million sites, enabling faster design of advanced quantum materials.
how weak-to-strong generalization works in neural networks
a new theory shows that fine-tuning a strong model on a weak model's outputs can elicit pre-trained knowledge without losing general skills.
online conformal prediction with nested sets
new methods produce nested prediction sets across multiple coverage levels for simultaneous uncertainty quantification in online settings.
learning transferable user preferences from limited interactions
a new framework learns latent user preferences from few conversations and applies them across tasks to improve human-aligned decision making.
verifier-guided action selection for embodied agents
a test-time framework uses a generative verifier to pick the best action from multiple candidates, improving robustness of multimodal language model agents without retraining.
quide: a single score for quantized model efficiency
a new metric called quide collapses compression, accuracy, and latency into one score to find optimal quantization levels for neural networks.
speech pauses and filler words may signal early dementia risk
ai analysis of natural speech patterns like pauses and filler words can predict executive function and may help detect early cognitive decline.
adaptive steering for discrete diffusion language models
uniform control schedules hurt text quality in discrete diffusion models, but a new adaptive scheduler improves multi-attribute steering by aligning interventions with each attribute's unique denoising timeline.
spatial adapter adds structured residuals to frozen models
a new post-hoc layer gives frozen predictors a spatial view of errors and a closed-form covariance without retraining.
valid inference after active data collection
a new framework enables valid statistical inference when reusing data from adaptive sampling methods like bayesian optimization.
adamw-trained transformers converge to ode system
large-depth transformers trained with adamw converge uniformly to a forward-backward ode system, with explicit convergence rates.
vertex-softmax tightens transformer verification
new method finds exact softmax bounds from score intervals, reducing slack in transformer robustness checks.
balora adds bayesian uncertainty to lora fine-tuning
balora extends lora with bayesian adaptation, improving accuracy and providing uncertainty estimates for large model fine-tuning.
decentralized novelty detection with quantized model exchange
a method for decentralized novelty detection that controls false discovery rate without sharing raw data, using low-precision model exchange.
capability elicitation vs creation in post-training
a free-energy perspective distinguishes whether post-training reweights existing model behaviors or creates new ones.
active multi-predictor inference cuts label costs
a new method routes each case to a cost-appropriate set of predictors and samples labels based on uncertainty, reducing variance under a fixed budget.