Research map / Video / Video-language understanding
Temporal and Cross-modal Alignment for Enhanced Audiovisual Video Captioning
Chen Zhao, Jiajun Ma, Qilong Huang, Tiehan Fan, Hongyu Li, Zhuoliang Kang, Xiaoming Wei, Jian Yang, Ying Tai
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 |
|---|---|---|---|
| fine-tune Qwen3-Omni with OCC-refined data (TCA-Captioner) | Qwen3-Omni-30B-A3B · TCA-Bench | AV Binding total: 42.5 -> 73.2 | Improved |
| fine-tune Qwen3-Omni with OCC-refined data (TCA-Captioner) | Qwen3-Omni-30B-A3B · TCA-Bench | AV Temporal total: 47.9 -> 76.9 | Improved |
| fine-tune Qwen3-Omni with OCC data | Qwen3-Omni · video-SALMONN-2 | Total error: 45.6 -> 31.5 | Improved |
| add Audio Checker to Observer data (ablation) | OCC pipeline · TCA-Bench | Audio: 57.6 -> 61.5 | Improved |
| add Visual Checker to Observer data (ablation) | OCC pipeline · TCA-Bench | Visual: 56.1 -> 64.8 | Improved |
| add A/V-Clip-Checker to full OCC (ablation) | OCC pipeline · TCA-Bench | AV Binding: 66.5 -> 73.2 | Improved |
| add A/V-Clip-Checker to full OCC (ablation) | OCC pipeline · TCA-Bench | AV Temporal: 69.8 -> 76.9 | Improved |
| add Audio Checker alone (ablation) | OCC pipeline · TCA-Bench | AV Temporal: 69.8 -> 68.1 | Hurt |
| add DPO with 2,000 OCC preference pairs (ablation) | Qwen3-Omni-30B SFT · TCA-Bench | AV Temporal: 76.9 -> 79.2 | Improved |
| add DPO with OCC preference pairs (ablation) | Qwen3-Omni-30B SFT · TCA-Bench | Visual: 63.6 -> 66.8 | Improved |
| use Qwen2.5-Omni instead of Qwen3-Omni as backbone (ablation) | TCA-Captioner SFT · TCA-Bench | AV Temporal: 76.9 -> 73.7 | Hurt |
| TCA-Captioner vs closed-source Gemini-3.0-Pro | TCA-Captioner · TCA-Bench | AV Binding total: 76.8 (Gemini-3.0-Pro) vs 73.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- AVoCaDO — AVoCaDO: An Audiovisual Video Captioner Driven by Temporal Orchestration (ICLR 2026) · better on 4 of 4 shared results
- video-SALMONN-2 — video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models (ICML 2024) · 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
- VideoComp: Advancing Fine-Grained Compositional and Temporal Alignment in Video-Text Models (CVPR 2025 · similarity 0.89)
- An eye for an ear: zero-shot audio description leveraging an image captioner with audio-visual token distribution matching (NeurIPS 2024 · similarity 0.89)
- OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning (NeurIPS 2026 · similarity 0.89)
- LongVALE: Vision-Audio-Language-Event Benchmark Towards Time-Aware Omni-Modal Perception of Long Videos (CVPR 2025 · similarity 0.88)
- VideoCon: Robust Video-Language Alignment via Contrast Captions (CVPR 2024 · similarity 0.88)
- LanguageBind: Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment (ICLR 2024 · similarity 0.88)
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.
