Research map / Video / Video-language understanding
DisCo: Towards Distinct and Coherent Visual Encapsulation in Video MLLMs
Jiahe Zhao, Rongkun Zheng, Yi Wang, Helin Wang, Hengshuang Zhao
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 |
|---|---|---|---|
| add DisCo encapsulation | InternVideo2-HD · MVBench | accuracy: 66.3 -> 68.2 | Improved |
| add DisCo encapsulation | InternVideo2 · STAR | accuracy: 64.5 -> 72.7 | Improved |
| add DisCo encapsulation | ST-LLM · MVBench | accuracy: 54.7 -> 58.0 | Improved |
| add DisCo encapsulation | InternVideo2-HD · VideoChatGPT-Bench | avg score: 2.96 -> 3.24 | Improved |
| add DisCo encapsulation | ST-LLM · VideoMME medium (w/o sub) | accuracy: 40.2 -> 39.6 | Hurt |
| fine-tune on same data without modules (SFT) (ablation) | InternVideo2-HD · MVBench | accuracy: 66.3 -> 66.5 | No clear effect |
| add VCD module (ablation) | InternVideo2-HD · EgoSchema subset | accuracy: 67.0 -> 71.8 | Improved |
| add TFC module (ablation) | InternVideo2-HD · EgoSchema subset | accuracy: 67.0 -> 70.4 | Improved |
| add TFC on top of VCD (full DisCo) (ablation) | InternVideo2-HD · EgoSchema subset | accuracy: 71.8 -> 72.2 | Improved |
| add global tokens to VCD (ablation) | DisCo · MVBench | accuracy: 65.8 -> 68.2 | Improved |
| use 16 groups instead of 4 (ablation) | DisCo VCD · MVBench | accuracy: 66.5 -> 68.2 | Improved |
| use 16 groups instead of 64 (ablation) | DisCo VCD · MVBench | accuracy: 67.3 -> 68.2 | Improved |
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.
- ST-LLM — ST-LLM: Large Language Models Are Effective Temporal Learners (ECCV 2024) · better on 3 of 3 shared results
- DeCo — DeCo: Decomposition and Reconstruction for Compositional Temporal Grounding via Coarse-To-Fine Contrastive Ranking (CVPR 2023) · 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
- Exploring the Design Space of Visual Context Representation in Video MLLMs (ICLR 2025 · similarity 0.90)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.90)
- Needle In A Video Haystack: A Scalable Synthetic Evaluator for Video MLLMs (ICLR 2025 · similarity 0.89)
- TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning (ICLR 2025 · similarity 0.89)
- MVBench: A Comprehensive Multi-modal Video Understanding Benchmark (CVPR 2024 · similarity 0.89)
- What Do Visual Tokens Really Encode? Uncovering Sparsity and Redundancy in Multimodal Large Language Models (CVPR 2026 · similarity 0.89)
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.
