Jack Bai

@jackbot_cs

Champaign, IL

Pinned
Jack Bai
@jackbot_cs
Oct 14
1/11 Can a single-token importance weight preserve the target of full-sequence correction? With Future-KL, we prove that both updates share the same unique stationary policy. We also contribute the promising empirical results we observed. arxiv.org/abs/2610.18738
3
Jack Bai
@jackbot_cs
Oct 30
I firmly believe that loop transformer is just another instance of full width transformer. It improves performance by replacing argmax with full-width connection. Any other perspectives?
Jack Bai
@jackbot_cs
Oct 16
Ratio-free off-policy RL algorithm that shares some common insights in proof with Proximal KL.
Xuanchang Zhang
@thunderous777
Oct 16
1/7 Can an LLM learn from an older rollout model without importance weights? Introducing Gibbs-Gap Policy Optimization (GGPO). Sync rollouts every 8 -> 128 optimizer updates: accuracy stays at 46.58% -> 46.39% across five math benchmarks (Qwen3-1.7B-Base).
Jack Bai
@jackbot_cs
Jun 29
So glad to see that WebGym is being used as an environment in training Qwen-AgentWorld!!
1
Jack Bai
@jackbot_cs
Jun 9
😈 WebGym is now 300% faster with better performance. We introduce AsyncWebRL, with 300% end-to-end throughput and better performance over WebGym (which uses REINFORCE) with async GRPO. Two algorithmic takeaways we learnt: (1) do asynchronous multi-step agentic RL with the
1

Jack Bai’s posts