arXiv:2607. 16813v1 Announce Type: new Abstract: Sparse-support uncertainty is usually quantified by treating the dictionary as known, an assumption that can produce overconfident, label-dependent conclusions when the dictionary is learned from latent sparse mixtures.
arXiv:2608. 20295v1 Announce Type: new Abstract: Sparse pursuit after dictionary learning can yield a precise atom support even when its physical interpretation is not justified by the calibration data, especially for highly coherent dictionaries where alternative calibration-compatible dictionaries may assign different physical meanings to the same selected support.
arXiv:2607. 18305v1 Announce Type: cross Abstract: Some limits on what language models know are not gaps in data coverage but structural properties of learning from text.
arXiv:2606. 02385v1 Announce Type: cross Abstract: Sparse Autoencoders (SAEs) have found success parsing neural representations into interpretable concepts, providing a basis for understanding and control.
arXiv:2609.05877v1 Announce Type: new Abstract: Selecting compact training sets for machine-learned interatomic potentials requires deciding whether to preserve structural diversity or target configu...
The paper introduces Resolution-Aware Experimental Design (RAED), a method that selects experiments by minimizing the expected size of the nonempty structural candidate set while controlling false-exclusion rates. RAED is shown to preserve expected ordering under a composite Blackwell comparison and is implemented via a learned score-based approach with finite-sample nuisance-average and positive-tail calibration. Experiments on subsurface-flow, fluvial, and methane-oxidation benchmarks demonstrate RAED’s ability to resolve structural ambiguities and provide finite-sample guarantees for tail-sensitive nuisance risk.
The paper presents a rate‑distortion framework for understanding factual hallucination in closed‑book question answering. It shows that even when a fact is observed, limited memory forces it to be stored approximately, leading to errors that can be bounded by a combination of compression distortion and missing coverage. The authors derive a theoretical lower bound on error and validate it with simulations and probes on modern language models.
arXiv:2608. 19171v1 Announce Type: new Abstract: Deep models for irregularly-sampled time series answer queries at arbitrary continuous timestamps, yet report nothing about how far each answer should be trusted.
arXiv:2608. 10869v1 Announce Type: new Abstract: Worst-case multiclass bounds do not become smaller when the best classifier is already nearly correct: what is missing is an optimistic rate, a guarantee whose fluctuation scales with the oracle risk itself.
arXiv:2609. 12259v1 Announce Type: new Abstract: Matrix-valued memories make rank the natural budget of a learned representation: the number of independent directions a state spans bounds what it can bind, compose, and track.
arXiv:2602. 16568v2 Announce Type: replace-cross Abstract: Sparse recovery is among the most well-studied problems in learning theory and high-dimensional statistics.
The paper introduces Resolution-Aware Experimental Design (RAED), a method that selects experiments by minimizing the expected size of the nonempty structural candidate set while controlling false exclusions. RAED is shown to align with a composite Blackwell comparison and is implemented via a learned score-based approach with finite-sample calibration. Experiments on subsurface-flow, fluvial, and methane-oxidation benchmarks demonstrate that RAED can diverge from expected-information-gain selections, yielding clearer resolution and explicit ambiguity handling.
arXiv:2609.07061v2 Announce Type: replace Abstract: Physics-Informed Neural Networks (PINNs) embed PDE residuals into neural network training, but their internal representations remain opaque: it is...