Research map / Video / Video-language understanding
R^2-Tuning: Efficient Image-to-Video Transfer Learning for Video Temporal Grounding
Ye Liu, Jixuan He, Wanhua Li, Junsik Kim, Donglai Wei, Hanspeter Pfister, Chang Wen Chen
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 video-level contrastive constraint (ablation) | R2-Tuning · QVHighlights val | MR mAP: 44.01 -> 46.74 | Improved |
| add layer-wise contrastive constraint (ablation) | R2-Tuning · QVHighlights val | MR mAP: 44.01 -> 44.60 | Improved |
| combine video-level and layer-wise constraints (ablation) | R2-Tuning · QVHighlights val | MR mAP: 44.01 -> 47.59 | Improved |
| refine features in reversed (back-to-front) order instead of forward (ablation) | R2-Tuning · QVHighlights val | MR mAP | Improved |
| fuse multi-layer CLIP features (mean of last K layers) instead of a single layer (ablation) | R2 Block · QVHighlights val | MR mAP | Improved |
| cross-attention before self-attention instead of self before cross (ablation) | R2 Block · QVHighlights val | MR mAP: 45.21 -> 47.59 | Improved |
| share R2 Block parameters across refinement steps (ablation) | R2-Tuning · QVHighlights val | MR mAP: 47.48 -> 47.59 | No clear effect |
| increase temporal transformer layers from 1 to 6 (ablation) | R2 Block · QVHighlights val | MR mAP: 47.59 -> 47.42 | No clear effect |
| use R2-Tuning instead of full fine-tuning of CLIP | CLIP ViT-B/32 · QVHighlights val | MR mAP: 42.41 -> 47.59 | Improved |
| full fine-tuning of CLIP versus frozen CLIP | CLIP ViT-B/32 · QVHighlights val | MR mAP: 42.22 -> 42.41 | No clear effect |
| use LoRA on CLIP instead of frozen CLIP | CLIP ViT-B/32 · QVHighlights val | MR mAP: 42.22 -> 41.29 | Hurt |
| use R2-Tuning without video backbone against UniVTG with SlowFast and post pre-training | QVHighlights test · QVHighlights test | MR mAP avg: 43.63 -> 46.17 | Improved |
Later papers that compared with this method
6 later accepted papers report R2-Tuning as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Object-Centric Framework for Video Moment Retrieval (AAAI 2026 · better on 2 of 3 shared results)
- HERO: Hierarchical Embedding-Refinement for Open-Vocabulary Temporal Sentence Grounding in Videos (CVPR 2026 · better on 2 of 2 shared results)
- Moment Quantization for Video Temporal Grounding (ICCV 2025 · better on 5 of 7 shared results)
- Sparse-Dense Side-Tuner for efficient Video Temporal Grounding (ICCV 2025 · better on 5 of 5 shared results)
- D&M: Enriching E-commerce Videos with Sound Effects by Key Moment Detection and SFX Matching (AAAI 2025 · better on 3 of 3 shared results)
- OVG-HQ: Online Video Grounding with Hybrid-modal Queries (ICCV 2025 · better on 2 of 2 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
- Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding (NeurIPS 2025 · similarity 0.90)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.88)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video Grounding (NeurIPS 2025 · similarity 0.88)
- Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos (CVPR 2025 · 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.
