AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Don't Let the Video Speak: Audio-Contrastive Preference Optimization for Audio-visual Language Models

Ami Baid, Zihui Xue, Kristen Grauman

ECCV 2026 · Vision-language and multimodal › Cross-modal learning · approach: Audio-visual speech separation · 4 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
apply ACPO instead of base Video-LLaMA2Video-LLaMA2-7B-AV · AVHBench audio hallucinationaccuracy: 74.0 -> 79.9Improved
apply ACPOVideo-LLaMA2-7B-AV · CMM Overreliance on Visionaccuracy on no-instances: 64.5 -> 82.0Improved
apply ACPOVideo-LLaMA2-7B-AV · audio captioning (original)CIDEr x100: 27.9 -> 43.6Improved
apply ACPOVideo-LLaMA2-7B-AV · audio captioning (swap)CIDEr x100: 13.1 -> 30.6Improved
apply ACPOVideo-LLaMA2-7B-AV · video captioning (original)METEOR: 22.7 -> 20.4Hurt
apply ACPO to other backbonesvideo-SALMONN-2 · AVHBench audio hallucinationaccuracy: 63.2 -> 68.8Improved
apply ACPO to Qwen2.5-OmniQwen2.5-Omni · AVHBench audio hallucinationaccuracy: 66.7 -> 69.3Improved
add audio-attribution pairs (ablation)ACPO · AVHBench audio hallucinationaccuracy: 79.1 -> 79.9Improved
add audio-sensitivity pairs (ablation)ACPO · AVHBench audio hallucinationF1: 79.6 -> 80.4Improved
use audio-swapped inputs for attribution pairs rather than no swap (ablation)ACPO attribution pairs · AVHBench audio hallucinationaccuracy: 76.9 -> 77.7Improved
train sensitivity pairs alone (no attribution pairs) (ablation)ACPO · AVHBench video hallucinationaccuracyHurt
use SFT on audio-swapped captions instead of preference optimization (ablation)Video-LLaMA2-7B-AV · AVHBench video hallucinationaccuracy: 77.2 -> 64.3Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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