AIMS: Intent-Aware Safety Classification Collection Human-annotated intent dataset and intent-aware safety classifiers (SFT, DPO, distillation, GRPO) for robust LLM guardrails. • 5 items • Updated Jun 23 • 6
Bootstrapping World Models from Dynamics Models in Multimodal Foundation Models Paper • 2506.06006 • Published Jun 6, 2025 • 15