When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models
Paper recorded by Signals 4 on 2026-10-27 in cs.CV. Abstract reproduced from arXiv; link to the original below.
Published 2026-10-27 on arXiv · recorded by Signals 4 on 2026-10-28
Category: cs.CV · 计算机视觉 · first seen 2026-10-28
Abstract
For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable q
Read on arXiv →
Cite this page: When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models: the #123 most recent of 739 cs.CV papers we have recorded (as of 2026-10-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/when-visual-evidence-is-insufficient-how-instructions-post-training-and-internal.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable:
papers.json
Get 4 AI signals a day by email — free.
Get 4 AI signals a day by email — free