Jiaming Ji
Hi there! I am a first-year PhD student at Institute for AI, Peking University, advised by Prof. Yaodong Yang (both a good teacher and a helpful friend in my life).
My research interests cover Safety Alignment and AI Safety. Hope you enjoy my works!
News
Feb, 2024 |
We released Aligner: a new efficient alignment paradigm, bypasses the whole RLHF process.
无需RLHF显著提升GPT-4/Llama2性能,北大团队提出Aligner对齐新范式 |
---|---|
Jan, 2024 |
Two papers get accepted to ICLR 2024!
Safe RLHF (Spotlight), SafeDreamer. |
Dec, 2023 |
One paper get accepted to JMLR 2023!
Heterogeneous-Agent Reinforcement Learning. |
Nov, 2023 |
One paper get accepted to TPAMI 2023!
Bi-DexHands: Towards Human-Level Bimanual Dexterous Manipulation. |
Nov, 2023 | Big News! We released AI Alignment: A Comprehensive Survey. |
Oct, 2023 |
We released Safe RLHF: Safe Reinforcement Learning from Human Feedback.
AK's Daily Papers |
Oct, 2023 | We released SafeDreamer: a novel algorithm for low-dimensional and vision-only safety tasks. |
Sep, 2023 |
I contributed to Baichuan’s model fine-tuning in RLHF as a core member and earned 1W+ ⭐.
Baichuan-7B Baichuan-13B Baichuan2 |
Sep, 2023 | Three papers get accepted to NeurIPS 2023! |
May, 2023 |
We released Safe-RLHF: Constrained Value Alignment for LLMs.
机器之心报道:国内首个可复现的RLHF基准,北大团队开源 PKU-Beaver |
Selected Publications
(*) indicates equal contribution2024
-
arXivRethinking Information Structures in RLHF: Reward Generalization from a Graph Theory PerspectiveIn Preprint, 2024
2023
-
VOCE: Variational Optimization with Conservative Estimation for Offline Safe Reinforcement LearningAdvances in Neural Information Processing Systems, 2023
-
AAAIAugmented proximal policy optimization for safe reinforcement learningProceedings of the AAAI Conference on Artificial Intelligence, 2023