AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Same Pool, Different Answer: Stable Best-of-N Selection for Vision-Language Models

Pengfei Zheng

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace length-norm selector with anchored selector (fixed reference, per-token average)BoN on LLaVA-1.5 7B · SRP stability (500 items)flip-rate (%): 44.3 -> 24.9Improved
replace length-norm selector with anchored selectorBoN on LLaVA-1.5 7B · POPEhallucination rate (%): 9.6 -> 6.2Improved
add epsilon-margin tie-break to length-normBoN + length-norm · POPEhallucination rate (%): 9.6 -> 9.8No clear effect
add epsilon-margin tie-break to length-normBoN + length-norm · SRP stability (500 items)flip-rate (%): 44.3 -> 37.1Improved
use per-token averaging instead of sequence-level summation (ablation)anchored selector · SRP stability (200 items, dropout)flip-rate (%): 40.8 -> 24.1Improved
use fixed reference instead of context-free self-anchoring (PMI) (ablation)anchored selector · SRP stability (200 items, dropout)flip-rate (%): 35.4 -> 24.1Improved
use fixed reference instead of co-moving reference (ablation)anchored selector · SRP stability (200 items, dropout)flip-rate (%): 33.6 -> 24.1Improved
add the three training-time filters (length, prompt-copy, object-existence) (ablation)DSPA · Object hallucination (judge-est.)object halluc. rate (%): 23.6 -> 6.2Improved
add the three training-time filters (ablation)DSPA · AMBERaccuracy: 77.5 -> 81.7Improved
use separately aligned RLAIF-V model instead of anchored selector on DSPA poollength-norm selector · SRP stability (500 items)flip-rate (%): 44.3 -> 35.1Improved
apply anchored selector at 13B scaleLLaVA-1.5 13B · SRP stability (avg)flip-rate (%): 38.5 -> 22.3Improved
apply anchored selector to Qwen2.5-VL-7B-InstructQwen2.5-VL-7B-Instruct · Object hallucination (judge-est., open-ended)object halluc. rate (%): 38.4 -> 8.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy