AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

DistFlow: A Fully Distributed RL Framework for Scalable and Efficient LLM Post-Training

Zhixin Wang, Jiaming Xu, Tianyi Zhou, Mingjun Zhang, Liming Liu, Jiarui Hu, Dian Yang, TongYu Wang, Ping Zhang, Jinlong Hou, Siyuan Feng, Yuan Qi and 1 more

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Local Cache to the baseline data pipeline (ablation)DistFlow · Ablation, batch size 2048speedup: 1.38x speedup at Batch Size=2048Improved
add Double Buffering with asynchronous reset (ablation)DistFlow · Ablation, batch size 2048speedup: 1.38x to 1.45xImproved
add Load Balancing (Constrained LPT) (ablation)DistFlow · Ablation, batch size 4096speedup: 1.38x to 1.58xImproved
add Load Balancing (Constrained LPT) (ablation)DistFlow · Ablation, batch size 1024speedupNo clear effect
add Distributed Data Buffer (ablation)DistFlow · Ablation, batch sizes 2048 and 4096speedup: peak speedup of 1.70x (Batch Size=2048) and 1.69x (Batch Size=4096)Improved
increase context length from 8k to 64kDistFlow vs verl, 7B · Long-context GRPOspeedup over verl: 1.48x at an 8k context length to 2.03x at 64kImproved
use fully distributed DistFlow instead of centralized verl, GRPO instead of PPODistFlow · End-to-end throughputspeedup over verl: 1.09x - 1.64x for PPO; up to 2.62x for GRPOImproved
replace verl with DistFlow for 32B GRPO trainingDistFlow · Convergence, DeepScaleR, 32 GPUstotal execution time: reducing total execution time by 21%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy