AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Asymmetric Prompt Weighting for Reinforcement Learning with Verifiable Rewards

Reinhard Heckel, Mahdi Soltanolkotabi, Christos Thrampoulidis

ICML 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 0 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
use asymmetric prompt weighting (Linear-R, Plateau-R, Sqrt-R) instead of symmetric (GRPO, RLOO, Uniform-R)policy optimization · TinyZero countdown (Qwen2.5-3B)Pass@1: 0.74 -> 0.8Improved
use Plateau-R instead of GRPOLlama-3.1-8B · GSM8K from scratchPass@1Improved
use asymmetric weighting instead of symmetricLlama-3.2-3B-Instruct · MATH post-SFTPass@1: 0.53-56No clear effect
use asymmetric weighting instead of symmetricDeepSeek-R1-Distill-Qwen-1.5B · DAPO-math post-SFTPass@1No clear effect
add non-zero weight for zero-success prompts (ablation)Plateau-R · TinyZero countdownPass@1Improved
use kimi-weighting (zero weight at rho=0) instead of Linear-R (ablation)asymmetric weighting · TinyZero countdownPass@1Hurt
use RLOO/GRPO gradient direction instead of REINFORCE direction (rejection sampling) (ablation)Linear-R · TinyZero countdownPass@1Improved
three independent runs of Linear-R (ablation)Linear-R · TinyZero countdownPass@1No clear effect

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