Research map / Reinforcement learning / RL from feedback
Asymmetric Prompt Weighting for Reinforcement Learning with Verifiable Rewards
Reinhard Heckel, Mahdi Soltanolkotabi, Christos Thrampoulidis
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 changed | Where | Measured | Effect |
|---|---|---|---|
| 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 -> 0.8 | Improved |
| use Plateau-R instead of GRPO | Llama-3.1-8B · GSM8K from scratch | Pass@1 | Improved |
| use asymmetric weighting instead of symmetric | Llama-3.2-3B-Instruct · MATH post-SFT | Pass@1: 0.53-56 | No clear effect |
| use asymmetric weighting instead of symmetric | DeepSeek-R1-Distill-Qwen-1.5B · DAPO-math post-SFT | Pass@1 | No clear effect |
| add non-zero weight for zero-success prompts (ablation) | Plateau-R · TinyZero countdown | Pass@1 | Improved |
| use kimi-weighting (zero weight at rho=0) instead of Linear-R (ablation) | asymmetric weighting · TinyZero countdown | Pass@1 | Hurt |
| use RLOO/GRPO gradient direction instead of REINFORCE direction (rejection sampling) (ablation) | Linear-R · TinyZero countdown | Pass@1 | Improved |
| three independent runs of Linear-R (ablation) | Linear-R · TinyZero countdown | Pass@1 | No clear effect |
Closest papers by meaning
- LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL (NeurIPS 2026 · similarity 0.90)
- Act Only When It Pays: Efficient Reinforcement Learning for LLM Reasoning via Selective Rollouts (NeurIPS 2025 · similarity 0.89)
- Asymmetric REINFORCE for off-Policy Reinforcement Learning: Balancing positive and negative rewards (NeurIPS 2025 · similarity 0.88)
- Conditional Advantage Estimation for Reinforcement Learning in Large Reasoning Models (ICLR 2026 · similarity 0.88)
- One-Way Policy Optimization for Self-Evolving LLMs (ICML 2026 · similarity 0.88)
- Multi-Step Likelihood-Ratio Correction for Reinforcement Learning with Verifiable Rewards (NeurIPS 2026 · similarity 0.88)
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.
