AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shaping

Thanh-Long V. Le, Myeongho Jeon, Kim Vu, Viet Dac Lai, Eunho Yang

ICLR 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 RL-ZVP entropy-guided advantage for zero-variance promptsGRPO · AIME25 (Qwen3-8B-Base, DAPO-Math-17k)Acc@8: 20.00 -> 24.58Improved
add RL-ZVP advantage for zero-variance prompts at equal rollout budget vs dynamic-sampling filteringGRPO-DS / GRESO · six math benchmarks, large scaleAcc@8: +19.20 Acc@8 avg over r baselinesImproved
use RL-ZVP instead of filtering baselines given 3-5x more rollouts (-g settings)GRPO-DS-g / GRESO-g · six math benchmarks, large scaleAcc@8: +5.51 Acc@8 avgImproved
add reward/penalty for all-incorrect (negative) zero-variance prompts (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 47.75 -> 49.90Improved
add reward for all-correct (positive) zero-variance prompts (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 47.25 -> 49.90Improved
add token-entropy scaling of advantage magnitude (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 46.88 -> 49.90Improved
set alpha to 0.10 instead of 0.05 or 0.30 (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 46.78 (0.05), 45.81 (0.30) -> 49.90 (0.10)Improved
increase alpha to 1.0 and 10.0 (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 49.90 -> 44.32 (1.0), 39.96 (10.0)Hurt
use symmetric (sign-flipped entropy) advantage for negative prompts (ablation)RL-ZVP · Minerva/AMC23/MATH500 averageAcc@8: 49.90 -> 46.64Hurt
train with RL-ZVP on NuminaMath-CoT subsetGRPO · Qwen3-1.7B-Base NuminaMath-CoT, six benchmarksAcc@8/Pass@8: Pass@8 AMC23 -1.39, AIME24 -0.27; other gains up to +4.27Mixed
apply RL-ZVP to Llama-3.2-3B-InstructGRPO · Llama-3.2-3B-Instruct trained on MATHAcc@8: +1.20 Minerva, +2.50 AMC23, +2.50 AIME24; 0.00 AIME25Improved
apply RL-ZVP trained on MATH to out-of-domain tasksGRPO · LiveCodeBench/MMLU-Pro/GPQA-Diamond/BBH averageAcc@8: 27.84 -> 30.19Improved

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