Jiaming Ji
Ph.D Student
Ph.D (2023), Peking University
Interests
- Reinforcement Learning
- Safety Alignment
- AI for Science
Latest
- AI Alignment: A Comprehensive Survey
- Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback
- SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning
- Generative RLHF-V: Learning Principles from Multi-modal Human Preference
- InterMT: Multi-Turn Interleaved Preference Alignment with Human Feedback
- ProgressGym: Alignment with a Millennium of Moral Progress
- Reward Generalization in RLHF: A Topological Perspective
- SAE-V: Interpreting Multimodal Models for Enhanced Alignment
- Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction
- Sequence to Sequence Reward Modeling: Improving RLHF by Language Feedback
- Align Anything: Training All-Modality Models to Follow Instructions with Language Feedback
- OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Research
- Aligner: Efficient Alignment by Learning to Correct
- SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset
- Language Models Resist Alignment: Evidence From Data Compression
- ProgressGym: Alignment with a Millennium of Moral Progress
- PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference
- Safe RLHF: Safe Reinforcement Learning from Human Feedback
- Safety-Gymnasium: A Unified Safe Reinforcement Learning Benchmark
- BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset