pith. sign in

arxiv: 2506.02355 · v2 · pith:EOQZGMUFnew · submitted 2025-06-03 · 💻 cs.LG

Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening

classification 💻 cs.LG
keywords grpomodelbiasrankdistributionprovingtheoremformal
0
0 comments X
read the original abstract

Reinforcement learning is emerging as a primary driver for improving language model reasoning capabilities. A fundamental question is whether current reinforcement learning algorithms -- such as Group Relative Policy Optimization (GRPO), the de facto standard algorithm used to improve language model reasoning -- merely sharpen the base model's distribution around problems it can already solve. We investigate this question in the context of formal theorem proving, which has access to a perfect verifier. We identify a degenerate rank bias in GRPO in which highly probable trajectories are reinforced and rare ones are neglected. This results in distribution sharpening: the model can solve some problems with fewer samples, but underperforms simply sampling more solutions from the original model. To overcome GRPO's rank bias we introduce unlikeliness reward, a simple method for explicitly up-weighting rare but correct solutions. We show that unlikeliness reward mitigates rank bias and improves pass@$N$ across a large range of $N$ in both synthetic and real theorem proving settings. We also uncover an unexpected link between rank bias and a seemingly mundane hyperparameter -- the number of updates per batch -- that leads to a second, complementary mitigation. We combine our insights into a revised GRPO training recipe for formal theorem proving, yielding an open pipeline that achieves competitive performance to DeepSeek-Prover-V1.5-RL on the miniF2F-test benchmark. We release our implementation at https://github.com/AndreHe02/rewarding-unlikely-release

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning

    cs.LG 2026-06 unverdicted novelty 6.0

    Introduces a hierarchical latent selection model showing SFT supplies raw module materials in compound traces while RL decomposes them to identify atomic modules and enable recombination for new reasoning configurations.

  2. On Advantage Estimates for Max@K Policy Gradients

    cs.LG 2026-06 unverdicted novelty 6.0

    Proposes MaxPO using a Leave-Two-Out baseline for centered unbiased advantages in max@K policy gradients, with a unified derivation of finite-batch estimators.

  3. Taming Extreme Tokens: Covariance-Aware GRPO with Gaussian-Kernel Advantage Reweighting

    cs.CL 2026-05 unverdicted novelty 6.0

    Covariance-weighted GRPO with Gaussian-kernel reweighting tames extreme tokens to stabilize training and boost reasoning performance over standard GRPO.

  4. Teacher-Free Self-Training Amplifies but Does Not Compound: A Pass@$K$ Crossover on a Free-Verifier Domain

    cs.LG 2026-06 unverdicted novelty 5.0

    Self-training on verified outputs in a free-verifier DSL amplifies performance at pass@8 but does not expand reach, as the base model overtakes at pass@64 across trajectories.

  5. Mid-Training with Self-Generated Data Improves Reinforcement Learning in Language Models

    cs.AI 2026-05 unverdicted novelty 5.0

    Mid-training LLMs on self-generated diverse reasoning paths improves subsequent RL performance on mathematical benchmarks and OOD tasks.

  6. Beyond Distribution Sharpening: The Importance of Task Rewards

    cs.LG 2026-04 unverdicted novelty 5.0

    Task-reward reinforcement learning yields robust gains on math benchmarks for models like Llama-3.2-3B while distribution sharpening alone delivers only limited and unstable improvements.

  7. Rethinking Entropy Interventions in RLVR: An Entropy Change Perspective

    cs.LG 2025-10 unverdicted novelty 5.0

    Derives a token-level entropy change approximation revealing four factors, identifies limitations in prior entropy interventions, and proposes STEER which adaptively reweights tokens to mitigate collapse and improve p...

  8. Polychromic Objectives for Reinforcement Learning

    cs.LG 2025-09 unverdicted novelty 5.0

    Introduces polychromic objectives adapted into PPO via vine sampling and modified advantages, showing higher success rates and better coverage under perturbations on BabyAI, Minigrid, and algorithmic tasks.

  9. BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewards

    cs.AI 2026-06 unverdicted novelty 4.0

    BV-Blend blends prompt-local and semantic-cluster historical reward statistics via SEM-derived weights to stabilize critic-free RL advantage estimation.