AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

QuRL: Low-Precision Reinforcement Learning for Efficient Reasoning

Yuhang Li, Reena Elangovan, Xin Dong, Priyadarshini Panda, Brucek Khailany

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
use decoupled/ACR QuRL instead of plain quantized RL (INT8)RL INT8 · GSM8K (PPO)accuracy: 48.78 -> 53.55Improved
use QuRL ACR instead of FlashRL TIS (INT8)FlashRL INT8 · GSM8K (PPO)accuracy: 51.40 -> 53.55Improved
use QuRL ACR instead of FlashRL TIS (FP8)FlashRL FP8 · GSM8K (PPO)accuracy: 53.60 -> 54.28Improved
use quantized FP8 rollout without decoupled correctionRL FP8 · GSM8K (PPO)accuracy: 55.35 -> 0.0Hurt
quantize rollout to INT8 with plain RLRL BF16 · AIME 2024 (DAPO)Avg@32: 31.67 -> 0.001Hurt
add UAQ (invariant scaling s=1.5) to QuRL INT8 (ablation)QuRL w/o UAQ INT8 · AIME 2024 (DAPO)Avg@32: 30.63 -> 31.25Improved
add UAQ to QuRL FP8 (ablation)QuRL w/o UAQ FP8 · AIME 2024 (DAPO)Avg@32: 33.12 -> 33.27No clear effect
add UAQ to QuRL INT8 (ablation)QuRL w/o UAQ INT8 · DeepScaleR 5-task averageAvg@32 average: 54.79 -> 55.48Improved
add QuRL with UAQ to INT8 RLRL INT8 · DeepScaleR 5-task averageAvg@32 average: 52.31 -> 55.48Improved
use QuRL w/ UAQ instead of FlashRL (INT8)FlashRL INT8 · DeepScaleR 5-task averageAvg@32 average: 53.80 -> 55.48Improved
raise UAQ scale from 1.5 to 2 (ablation)QuRL INT8 · AIME 2024 (DAPO INT8)Avg@32: 31.25 -> 29.15Hurt
increase learning rate instead of UAQ scale (ablation)QuRL INT8 s=1 · AIME 2024 (DAPO INT8)Avg@32: 30.63 -> 26.66Hurt

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