Research map / Video / Video-language understanding
AV-Reasoner: Improving and Benchmarking Clue-Grounded Audio-Visual Counting for MLLMs
Lidong Lu, Guo Chen, Zhu Wei, Zhiqi Li, Yicheng Liu, Tong Lu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| train with GRPO curriculum (AV-Reasoner) instead of base | Ola-Omni-7B · DVD-Counting | accuracy: 16.50 -> 43.50 | Improved |
| train with GRPO curriculum | Ola-Omni-7B · CG-AV-Counting (Long Acc) | accuracy: 17.92 -> 22.30 | Improved |
| add SFT only (ablation) | Ola-Omni-7B · DVD-Counting | accuracy: 16.50 -> 41.50 | Improved |
| add SFT only (ablation) | Ola-Omni-7B · CG-AV-Counting (Long Acc) | accuracy: 17.92 -> 15.00 | Hurt |
| add QA-based RL after SFT (ablation) | SFT model · DVD-Counting | accuracy: 41.50 -> 23.00 | Hurt |
| add grounding RL (ablation) | SFT + QA RL · CG-AV-Counting (Long Acc) | accuracy: 16.55 -> 18.21 | Improved |
| add counting RL (ablation) | SFT + QA + grounding RL · CG-AV-Counting (Long Acc) | accuracy: 18.21 -> 20.84 | Improved |
| add full-task RL (ablation) | CB-RL with SRM · CG-AV-Counting (Long Acc) | accuracy: 20.84 -> 21.03 | Improved |
| curriculum RL instead of GRPO on all data at once (ablation) | SFT model · CG-AV-Counting (Long Acc) | accuracy: 10.42 -> 20.84 | Improved |
| add Stage Review Mechanism (ablation) | CB-RL + FT-RL · CG-AV-Counting (Long Acc) | accuracy: 20.74 -> 21.03 | No clear effect |
| add Stage Review Mechanism (ablation) | CB-RL + FT-RL · LLP segment-level | accuracy: 61.30 -> 64.70 | Mixed |
| enable explicit thinking output at inference (ablation) | AV-Reasoner · AVHBench A2V | accuracy: 84.45 -> 82.45 | Hurt |
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.
- PAVE — PAVE: Patching and Adapting Video Large Language Models (CVPR 2025) · better on 1 of 1 shared result
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
- CG-Bench: Clue-grounded Question Answering Benchmark for Long Video Understanding (ICLR 2025 · similarity 0.89)
- OmniVideoBench: Towards Audio-Visual Understanding Evaluation for Omni MLLMs (ICLR 2026 · similarity 0.88)
- UNICBench: UNIfied Counting Benchmark for MLLM (CVPR 2026 · similarity 0.87)
- video-SALMONN-o1: Reasoning-enhanced Audio-visual Large Language Model (ICML 2025 · similarity 0.87)
- Mellow: a small audio language model for reasoning (NeurIPS 2025 · similarity 0.87)
- When Eyes and Ears Disagree: Can MLLMs Discern Audio-Visual Confusion? (AAAI 2026 · similarity 0.86)
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.
