I often write some unofficial manuscripts and slides to help collect my thoughts.

I maintain a resource list for GFlowNets here.

I maintain a (possibly outdated) paper list for out-of-distribution generalization here (thanks for Irina’s help!).

Blogs

Knowledge Flow: Scaling Reasoning Beyond the Context Limit
Yufan Zhuang, Liyuan Liu, Dinghuai Zhang, Chandan Singh, Yelong Shen, Jingbo Shang, Jianfeng Gao

FlashRL: 8Bit Rollouts, Full Power RL
Your Efficient RL Framework Secretly Brings You Off-Policy RL Training
Feng Yao*, Liyuan Liu*, Dinghuai Zhang, Chengyu Dong, Jingbo Shang, Jianfeng Gao
Blogs on addressing off-policy mismatch (from vLLM serving & rollout quantization) in modern LLM+RL systems

Towards 131k-Context dLLMs
Albert Ge, Chandan Singh, Dinghuai Zhang, Letian Peng, Yufan Zhuang, Ning Shang, Li Lyna Zhang, Liyuan Liu, Jianfeng Gao

Talk Slides

• At the intersection of probabilistic inference and exploration methods: link
• Review of counterfactual representation learning: link
• Review of out-of-distribution generalization: link
• Introduction to causal inference: link
• Review of learning in traditional cv methods: link
• Review of reweight methods: link
• Review of unlabeled data used in adversarial training: link
• Some random papers sharing: link
• One paper about off-policy evaluation: link
• Review of normalizing flows: link
• Review of capsule networks: link

Notes

• Bayesian optimal experimental design: link
• Kernelized Wasserstein gradient flow: link
• Conformal inference basics: link
• A way to unify different probabilistic inference approaches: link
• Connection between adversarial robustness and other learning fields: link
• Stochastic analysis: link
• A review on interpretating deep neural network: link
• A short survey on Image Inpainting: link