AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Towards High Data Efficiency in Reinforcement Learning with Verifiable Reward

Xinyu Tang, Zhenduo Zhang, Yurou Liu, Xin Zhao, zujie wen, Zhiqiang Zhang, JUN ZHOU

ICLR 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 12 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 PageRank-weighted DPP pruning (ablation)DEPO (DeepSeek-R1-Distill-Qwen-7B) · AIME25accuracy: 50.0 -> 50.9Improved
add difficulty-aware normal-distribution sampling (ablation)DEPO (DeepSeek-R1-Distill-Qwen-7B) · AIME25accuracy: 47.8 -> 50.9Improved
add explorability-based rollout selection instead of random (ablation)DEPO (DeepSeek-R1-Distill-Qwen-7B) · AIME25accuracy: 45.3 -> 50.9Improved
add replay of under-explored samples (ablation)DEPO (DeepSeek-R1-Distill-Qwen-7B) · AIME25accuracy: 48.4 -> 50.9Improved
add replay of under-explored samples (ablation)DEPO (DeepSeek-R1-Distill-Qwen-7B) · MATH500accuracy: 95.2 -> 95.9Improved
use DEPO offline selection instead of LearnalignGRPO on 20% DAPO-Math (DeepSeek-R1-Distill-Qwen-7B) · five-benchmark averageaverage accuracy: 58.7 -> 61.4Improved
use explorability rollout pruning instead of GRESODEPO offline subset (DeepSeek-R1-Distill-Qwen-7B) · five-benchmark averageaverage accuracy: 58.1 -> 61.1Improved
use random rollout pruning (40% rollouts) instead of full rolloutsDEPO offline subset (DeepSeek-R1-Distill-Qwen-7B) · five-benchmark averageaverage accuracy: 61.4 -> 56.7Hurt
online DEPO pruning vs full rollouts on DEPO subsetDeepSeek-R1-Distill-Qwen-7B · five-benchmark averageaverage accuracy: 61.4 -> 61.1No clear effect
DEPO offline on 20% data vs full dataset trainingDeepSeek-R1-Distill-Llama-8B · five-benchmark averageaverage accuracy: 58.2 -> 57.9No clear effect
use mean difficulty 0.75 instead of medium (ablation)difficulty-aware sampling · convergence performanceaccuracyHurt
use mean difficulty 0.25 instead of medium (ablation)difficulty-aware sampling · final performanceaccuracyMixed

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