AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Beyond Logits: Coherent Hallucination Mitigation via Attention Contrastive Decoding

Yujia Chen, Rui Sun, Huayu Mai, Wangkai Li, Zhangyu He, Bingzhou Wang, Aibing Li, Wenzhang SUN, Tianzhu Zhang

ICML 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Decoding-time hallucination mitigation

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes attention-level contrastive decoding for LVLMs as an alternative to logit-level contrastive decoding, aiming to reduce hallucination while preserving generation coherence.

Problem

LVLMs hallucinate factually inaccurate content; existing contrastive decoding mitigates this by contrasting amateur and expert branches at the logit level, which the authors argue compromises generation coherence and requires restrictive penalty constraints.

Approach

Moves the contrastive intervention from output logits to attention, contrasting expert and amateur attention so that hallucination mitigation does not damage coherence (details beyond the abstract not available).

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add ASS (adaptive subtraction) to attention contrastive decoding (ablation)VCD+ACD on LLaVA-1.5 · CHAIRCHAIRi: 15.6 -> 14.3Improved
add ASS to attention contrastive decoding (ablation)VCD+ACD on LLaVA-1.5 · CHAIRCHAIRs: 52.1 -> 51.0Improved
replace VCD logit penalty with uniform attention subtraction (ablation)VCD on LLaVA-1.5 · CHAIRCHAIRs: 51.8 vs 52.1No clear effect
add the penalty term to VCD (ablation)VCD on LLaVA-1.5 · CHAIRCHAIRs: 53.4 -> 51.8Improved
add ACD to VCDVCD on LLaVA-1.5 · CHAIRCHAIRi: 16.2 -> 14.3Improved
add ACD to ICDICD on LLaVA-1.5 · CHAIRCHAIRi: 15.5 -> 13.9Improved
add ACD to SIDSID on LLaVA-1.5 · CHAIRCHAIRi: 14.1 -> 13.3Improved
add ACD to VCDVCD on Qwen-VL · CHAIRCHAIRi: 13.7 -> 12.5Improved
add ACD to VCDVCD on LLaVA-1.5 · LLaVA-Bench-Wildtotal accuracy: 4.30 -> 6.10Improved
add ACD to VCDVCD on Qwen-VL · LLaVA-Bench-Wildtotal accuracy: 5.40 -> 7.11Improved
add ACD to VCDVCD on LLaVA-1.5 · POPE RandomF1: 86.7 -> 87.2Improved
add ACD to VCDVCD on LLaVA-1.5 · MMEPerception score: 1475.4 -> 1497.7Improved

Surprising findings

Experimental setup

Baselines: contrastive decoding

Builds on: contrastive decoding

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