AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

FlowRL: Matching Reward Distributions for LLM Reasoning

Xuekai Zhu, Daixuan Cheng, Dinghuai Zhang, Hengli Li, Kaiyan Zhang, Che Jiang, Youbang Sun, Ermo Hua, Yuxin Zuo, Xingtai Lv, Qizheng Zhang, Lin Chen and 11 more

ICLR 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 31 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 importance sampling to FlowRL (ablation)FlowRL · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 26.71 -> 35.63Improved
use trajectory-level IS instead of combined GFlowNets+PPO loss (ablation)FlowRL · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 33.67 -> 35.63Improved
raise beta from 5 to 15 (ablation)FlowRL · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 31.34 -> 35.63Improved
raise beta from 15 to 30 (ablation)FlowRL · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 35.63 -> 35.09Hurt
raise beta from 10 to 15 (ablation)FlowRL · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 34.41 -> 35.63Improved
replace GRPO with FlowRLGRPO · Math average, Qwen2.5-32B-BaseAvg@16 accuracy: 38.34 -> 48.39Improved
replace GRPO with FlowRLGRPO · Math average (Avg@16), Qwen2.5-7B-BaseAvg@16 accuracy: 32.48 -> 35.63Improved
replace PPO with FlowRLPPO · Math average, Qwen2.5-32B-BaseAvg@16 accuracy: 43.25 -> 48.39Improved
replace GRPO with FlowRLGRPO · LiveCodeBenchAvg@16: 32.75 -> 37.43Improved
replace GRPO with FlowRLGRPO · CodeForcesrating: 1313.82 -> 1549.47Improved
replace GRPO with FlowRLGRPO · HumanEval+Avg@16: 80.13 -> 83.28Improved
replace GRPO with FlowRL, temperature 1.0GRPO · Math average, Qwen2.5-7B (Avg@64, T=1.0)Avg@64 accuracy: 32.44 -> 34.62Improved

Later papers that compared with this method

6 later accepted papers report FlowRL as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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