Mind the Spike: Mechanisms and Brittleness of Visual Massive Activations in Large Vision-Language Models

Beyond Accuracy: Counterfactual Fragility and Demographic Bias in Clinical Evaluation of LLMs

Clinical LLM evaluation often emphasizes answer accuracy; however, accuracy alone does not test counterfactual consistency or demographic robustness.

2026-09-26 ·arxiv.org

Decision-Sufficient State Representations: Measuring and Reducing Write-Time Regret

Long tasks produce more history than an LLM agent can hold in its context, and more than it uses reliably even when the history fits.

2026-09-26 ·arxiv.org

Finding Emotions Where They Belong: Rethinking Audio Emotion Recognition through Masked Temporal Affective Grounding

Audio emotion recognition (AER) typically assigns a single label to an entire recording, leaving the temporal scope of that label ambiguous when multiple speakers and affective events are present.

2026-09-26 ·arxiv.org

Overwhelmed by Choice: Studying LLM Decision Making at Scale

Multiple-choice and candidate-selection evaluations are widely used to assess LLM reasoning and decision-making, yet most benchmarks contain relatively small candidate sets.

2026-09-26 ·arxiv.org

A Brief Perspective on Deep Learning Using Common Lisp

2026-09-26 ·youtube.com