AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

PopCD: Test-Time Behavior Enhancement via Polarity-Prompt Contrastive Decoding

Baolong Bi, Yuyao Ge, Shenghua Liu, Yuchen He, S.L. Tong, Lizhe Chen, Lingrui Mei, Zehao Li, Yiwei Wang, Yujun Cai, Ming-Hsuan Yang, Xueqi Cheng

TPAMI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Reinforcement learning for visual reasoning · 5 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
add the Adaptive Plausibility Constraint to PromptCD (ablation)PromptCD (LLaMA2-7B) · NQHit Rate: 30.7 -> 77.8Improved
add the Adaptive Plausibility Constraint to PromptCD (ablation)PromptCD (Qwen2.5-7B) · NQHit Rate: 62.4 -> 89.8Improved
set contrastive coefficient gamma to 0.5 instead of 0.2 (ablation)PromptCD (LLaMA2-7B-chat) · Helpfulness ConRConR: 77.32 -> 78.89Improved
increase gamma from 0.5 to 0.8 (ablation)PromptCD (LLaMA2-7B-chat) · Helpfulness ConRConR: 78.89 -> 70.53Hurt
increase gamma from 0.5 to 0.8 (ablation)PromptCD (LLaMA2-7B-chat) · Harmlessness DG-AVGDG-AVG: 89.98 -> 85.34Hurt
apply contrast on layers [20,25] instead of [10,15] (ablation)PromptCD (LLaVA-1.5-7B) · TextVQAaccuracy: 49.2 -> 58.2Improved
apply PromptCD contrastive decoding instead of vanilla decodingLLaMA2-7B-chat · NQConR: 43.35 -> 75.99Improved
apply PromptCD instead of vanilla decodingLLaMA2-7B-chat · SafeEditDefense Success: 40.59 -> 96.67Improved
apply PromptCD instead of DINMLLaMA3-8B-instruct · SafeEditDefense Success: 99.26 -> 95.26Hurt
apply PromptCD instead of vanilla decodingLLaVA-1.5-13B · V*accuracy: 42.4 -> 70.0Improved
apply PromptCD on top of MeLLoLLaMA2-7B-chat · MQuAKE-3kaccuracy: 32.6 -> 43.1Improved
apply PromptCD on top of IKELLaMA2-7B-chat · MQuAKE-2002accuracy: 20.6 -> 20.4No clear effect

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