Research map / Video / Video generation
LaMoGen: Language to Motion Generation Through LLM-Guided Symbolic Inference
Junkun Jiang, Ho Yin Au, Jingyu Xiang, Jie 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 GPT-4.1 LLM composer of conceptual Laban symbols (ablation) | LaMoGen · Laban Benchmark | FID: 5.562 -> 1.861 | Improved |
| add Qwen3 composer (ablation) | LaMoGen · Laban Benchmark | FID: 5.562 -> 1.903 | Improved |
| replace GPT-4.1 with human (ground-truth) concept cues (ablation) | LaMoGen · Laban Benchmark | FID: 1.861 -> 1.769 | Improved |
| use stronger LLM (GPT4.1 vs GPT4.1mini) as composer (ablation) | LaMoGen · HumanML3D test | FID: 0.561 -> 0.252 | Improved |
| add GPT-4.1 composer (ablation) | LaMoGen · HumanML3D test | R-precision Top-1: 0.438 -> 0.491 | Improved |
| use stronger LLM (GPT4.1 vs GPT4.1mini) as composer (ablation) | LaMoGen · KIT-ML test | FID: 0.550 -> 0.415 | Improved |
| replace rule-based hardcoded Laban detection with proposed detection workflow | Laban symbol detection · Laban-annotated walking motions | avg. SMT: 0.751 -> 0.871 | Improved |
| replace hardcoded detection with proposed workflow | Laban symbol detection · Laban-annotated walking motions | avg. HMN: 0.611 -> 0.786 | Improved |
| increase retrieved examples from 1 to 3 (ablation) | LaMoGen (GPT-4.1) · HumanML3D test | generation metrics | Improved |
| increase retrieved examples from 3 to 5 or 7 (ablation) | LaMoGen (GPT-4.1) · HumanML3D test | generation metrics | No clear effect |
| set masking ratio on Laban codes to 0.3 (ablation) | LaMoGen (GPT-4.1, top-3) · HumanML3D test | generation metrics | Improved |
| use LaMoGen (GPT4.1) instead of MotionGPT | MotionGPT · Laban Benchmark | SMT supL: 0.488 -> 0.583 | Improved |
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.
- CoMo — CoMo: Controllable Motion Generation through Language Guided Pose Code Editing (ECCV 2024) · better on 3 of 3 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
- MotionMaster: Generalizable Text-Driven Motion Generation and Editing (CVPR 2026 · similarity 0.91)
- LLaMo: Scaling Pretrained Language Models for Unified Motion Understanding and Generation with Continuous Autoregressive Tokens (CVPR 2026 · similarity 0.89)
- Fg-T2M++: LLMs-Augmented Fine-Grained Text Driven Human Motion Generation (IJCV 2025 · similarity 0.89)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · similarity 0.88)
- AMD: Anatomical Motion Diffusion with Interpretable Motion Decomposition and Fusion (AAAI 2024 · similarity 0.88)
- UniMo: Unified Motion Generation and Understanding with Chain of Thought (AAAI 2026 · similarity 0.87)
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.
