AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

DRIVE: Best Data Scheduling Practices for Reinforcement Learning with Verifiable Reward in Competitive Code Generation

Speed Zhu, Chuheng Zhang, Jianwei Cai, Guang Chen, Lulu Wu, Xiaolong Xu, Xuyun Zhang, Saiyong Yang, Wiggin Zhou

ICML 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards

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 two-stage RL (entropy expansion then progressive curriculum) after SFTQwen2.5-32B SFT model · Codeforces Weekly (33)pass@1: 0.115 -> 0.182Improved
add stage 1 entropy expansion before stage 2 (vs stage 2 only, 32k) (ablation)two-stage RL · LiveCode 08-11 (166)pass@1: 0.636 -> 0.699Improved
add stage 2 hard-sample curriculum after stage 1 (vs stage 1 only) (ablation)two-stage RL · Codeforces Weekly (33)pass@1: 0.112 -> 0.182Improved
use 24k instead of 32k max sequence length in stage 1 (ablation)RL stage 1 · LiveCode 08-11 (166)pass@1: 0.676 (32k) vs 0.625 (24k)Mixed
stage 2 trained on hard samples only rather than mixed difficulty (ablation)RL stage 2 · LiveCode 08-11 (166)pass@1Improved
use harder but fewer samples (109 internal) in stage 2 (ablation)RL stage 2 · LiveCode V6 (175)pass@1: 0.703 -> 0.743Improved
stage 2 with 24k max length instead of 32k (ablation)RL only stage 2 · LeetCode Weekly (32)pass@1: 0.550 -> 0.296Hurt
add arena learning data filtering (1.27M to 470k prompts) (ablation)SFT · LeetCode Weekly (32)pass@1: 0.558 -> 0.553No clear effect
double weight of hard samples via classifier on arena-filtered data (ablation)SFT · LeetCode Weekly (32)pass@1: 0.553 -> 0.578Improved
double weight of hard samples via classifier on arena-filtered data (ablation)SFT · LiveCode V5 (167)pass@1: 0.603 (basic) -> 0.594 (ours)Mixed
increase rollouts per prompt from 8 to 16 and 64 (ablation)single-sample RL training · single hard sampleaccuracyImproved
train on four hard samples instead of the single sample alone (ablation)RL · single hard sampleaccuracyImproved

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