AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Squeeze the Soaked Sponge: Efficient Off-policy RFT for Large Language Model

Jing Liang, Jinyi Liu, Yi Ma, Hongyao Tang, YAN ZHENG, Shuyue Hu, LEI BAI, Jianye HAO

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning

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 increased UTD ratio (ablation)ReMix-PPO · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1: 50.63 -> 52.10Improved
add KL-Convex policy constraint (ablation)ReMix-PPO · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1: 49.36 -> 52.10Improved
add policy reincarnation (ablation)ReMix-PPO · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1: 47.26 -> 52.10Improved
add UTD, KL-Convex and reincarnation on top of Mix-PPG (ablation)Mix-PPG · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1: 48.70 -> 52.10Improved
replace on-policy PPO with ReMix-PPO (1.5B, rollout efficiency)PPO · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1 at matched score: 50.61 at 0.230M -> 50.61 at 0.020MImproved
replace PPO (200 steps) with ReMix-PPO (75 steps)PPO · Math avg Pass@1 (5 benchmarks, 7B)avg Pass@1: 60.34 -> 64.39Improved
apply ReMix to GRPOGRPO · Math avg Pass@1 (5 benchmarks, 1.5B)avg Pass@1 (200 steps): 48.64 -> 50.53Improved
apply ReMix-PPO to code generation (1.5B)PPO · LiveCodeBenchPass@1: 19.35 -> 24.01Improved
apply ReMix-PPO to code generation (7B)PPO · LiveCodeBenchPass@1: 33.33 -> 34.41Improved
ReMix-PPO training on 1.5BR1-Distill-Qwen-1.5B · Math avg Pass@8Pass@8: 37.68 -> 51.72Improved
ReMix-PPO training on 7BR1-Distill-Qwen-7B · Math avg Pass@8Pass@8: 68.63 -> 75.94Improved
vary reincarnation step T from 50 to 25 (ablation)ReMix-PPO · Math avg Pass@1 (1.5B)avg Pass@1: 51.52 -> 49.29Hurt

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