SafeMo: Linguistically Grounded Unlearning for Trustworthy Text-to-Motion Generation

ACL ARR 2026 January Submission3491 Authors

Keywords: Machine Unlearning, Text-to-Motion Generation, Large Language Models, Safety Alignment

Abstract: Text-to-motion (T2M) generation have achieved impressive realism but pose significant safety risks by faithfully executing harmful textual prompts. Existing safety measures face dual challenges: brittle keyword filtering and discrete codebook manipulation, which degrade benign generation quality and introduce jerky transitions. To address these challenges, we propose SafeMo, a linguistically grounded framework that leverages Large Language Models (LLMs) to align motion generation with safety constraints. First, we introduce SafeMoEngine, an LLM-agent pipeline that autonomously classifies harmful intents and performs semantic rewriting to construct SafeMoVAE-29K, the first safety-aligned text-to-motion dataset. Second, we propose Minimal Motion Unlearning (MMU), a continuous-space unlearning strategy that projects harmful semantic concepts into a task vector for precise negation, avoiding the quantization artifacts of discrete methods. Experiments demonstrate effective unlearning performance of SafeMo by showing strengthened forgetting on unsafe prompts, reaching 2.5$\times$ and 14.4$\times$ higher forget-set FID on HumanML3D and Motion-X respectively, compared to the previous SOTA human motion unlearning method LCR, with benign performance on safe prompts being better or comparable.

Paper Type: Long

Research Area: Multimodality and Language Grounding to Vision, Robotics and Beyond

Research Area Keywords: cross-modal content generation, multimodality, safety and alignment, LLM/AI agents

Contribution Types: Publicly available software and/or pre-trained models, Data resources

Languages Studied: English

Submission Number: 3491