Research map / Video / Video generation
Chronologically Accurate Retrieval for Temporal Grounding of Motion-Language Models
Kent Fujiwara, Mikihiro Tanaka, Qing Yu
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 shuffled-event texts as hard negatives in contrastive training | TMR (DistilBERT) · HumanML3D motion-to-text retrieval, orig -> event | CAR accuracy: 64.81 -> 99.33 | Improved |
| add shuffled-event hard negatives | TMR (DistilBERT) · HumanML3D, event -> event | R@1: 6.93 -> 8.90 | Improved |
| remove the VAE text encoder when fine-tuning with negatives (ablation) | Ours t5-large · HumanML3D text-to-motion retrieval | R@1: 6.57 -> 8.03 | Improved |
| keep VAE but remove motion reconstruction decoder (ablation) | Ours t5-large · HumanML3D text-to-motion retrieval | R@1: 6.57 -> 4.65 | Hurt |
| fine-tune CLIP text encoder via TMR without negatives (ablation) | T2M-GPT · HumanML3D generation | R-Precision Top-1: 0.489 -> 0.498 | Improved |
| add chronological negatives on top of tuned CLIP (ablation) | T2M-GPT · HumanML3D generation | FID: 0.107 -> 0.070 | Improved |
| add chronological negatives on top of tuned CLIP (ablation) | ReMoDiffuse · HumanML3D generation | R-Precision Top-1: 0.525 -> 0.525 | No clear effect |
| train text encoder with chronological negatives | T2M-GPT · HumanML3D likelihood of true vs shuffled text | fraction preferring true text: 61.9% -> 89.9% | Improved |
| larger language model when fine-tuning (ablation) | Motion-language retrieval · HumanML3D text-to-motion retrieval | R@1: t5-large 8.03 vs CLIP 5.57 best-config | Improved |
Closest papers by meaning
- Beyond Global Alignment: Fine-Grained Motion-Language Retrieval via Pyramidal Shapley-Taylor Learning (ICML 2026 · similarity 0.90)
- Text Motion Translator: A Bi-Directional Model for Enhanced 3D Human Motion Generation from Open-Vocabulary Descriptions (ECCV 2024 · similarity 0.90)
- LaMP: Language-Motion Pretraining for Motion Generation, Retrieval, and Captioning (ICLR 2025 · similarity 0.89)
- AToM: Aligning Text-to-Motion Model at Event-Level with GPT-4Vision Reward (CVPR 2025 · similarity 0.89)
- VimoRAG: Video-based Retrieval-augmented 3D Motion Generation for Motion Language Models (NeurIPS 2025 · similarity 0.89)
- ReMoGPT: Part-Level Retrieval-Augmented Motion-Language Models (AAAI 2025 · 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.
