Towards Continual Motion-Language Agents: LoRA Variants for Incremental Motion Understanding and Generation
cs.LG, cs.AI
Submitted: 2026-06-29
Updated: 2026-11-02
Comments: 19 pages, 1 figure. Updated author version of the paper published at CoLLAs 2026. Adds merged-adapter analysis and comparative evaluations, clarifies routing and interpretation of results, and integrates the approach overview into the main text
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Parameter-Efficient Continual Fine-Tuning: A Survey
- Self-Supervised Training Enhances Online Continual Learning
- Continual Learning for Generative AI: From LLMs to MLLMs and Beyond
- MotionGPT: Human Motion as a Foreign Language
- Motion Generation: A Survey of Generative Approaches and Benchmarks
- Towards General Text Embeddings with Multi-stage Contrastive Learning
- InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learning
- An Empirical Investigation of the Role of Pre-training in Lifelong Learning
- Gemma: Open Models Based on Gemini Research and Technology
- MotionGPT-2: A General-Purpose Motion-Language Model for Motion Generation and Understanding
- Orthogonal Subspace Learning for Language Model Continual Learning
- Motion-Agent: A Conversational Framework for Human Motion Generation with LLMs
- MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Model
- Lifelong Learning of Large Language Model based Agents: A Roadmap
- iManip: Skill-Incremental Learning for Robotic Manipulation
- MotionGPT3: Human Motion as a Second Modality
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis