Paper

Do Vision-Language Models See or Guess? Measuring and Reducing Textual-Prior Reliance with a Phrasing-Controlled Benchmark

Unreadunread

AI summary

arXiv:2606. 10400v1 Announce Type: new Abstract: Vision-language models (VLMs) are increasingly deployed where answers must follow from what is in the image, yet they often answer from textual priors, the question's phrasing together with memorized world knowledge, rather than from the image itself, which inflates benchmark scores and yields confident but ungrounded answers.

Reading status

Priority

About this paper

Authors: Pratham Singla, Shivank Garg, Vihan Singh, Paras Chopra
Released: 2026-06-10 (29d ago)
Topic: Related to current project
Relation to my work: Read next because Do Vision-Language Models See or Guess? Measuring and Reducing Textual-Prior Reliance with a Phrasing-Controlled Benchmark overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, source, rate, control, leakage, language, model. Source: arxiv cs.CL (NLP).
arXiv: 2606.10400

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to Do Vision-Language Models See or Guess? Measuring and Reducing Textual-Prior Reliance with a Phrasing-Controlled Benchmark. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.