MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning

TL;DR

MASPO introduces a unified RLVR framework for LLMs that enhances gradient use, balances probability mass, and aligns signal reliability, leading to improved robustness and efficiency.

Contribution

It proposes MASPO, a novel framework combining differentiable gating, adaptive limiting, and asymmetric risk control for better LLM reinforcement learning.

Findings

MASPO outperforms existing methods in robustness and sample efficiency.

The framework effectively balances exploration and exploitation in LLM training.

Code is publicly available at https://github.com/FlyTune/MASPO-RL.

Abstract

Existing Reinforcement Learning with Verifiable Rewards (RLVR) algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanisms that are fundamentally misaligned with the complex optimization dynamics of Large Language Models (LLMs). In this paper, we identify three critical challenges in these methods: (1) inefficient gradient utilization caused by the binary cutoff of hard clipping, (2) insensitive probability mass arising from uniform ratio constraints that ignore the token distribution, and (3) asymmetric signal reliability stemming from the disparate credit assignment ambiguity between positive and negative samples. To bridge these gaps, we propose Mass-Adaptive Soft Policy Optimization (MASPO), a unified framework designed to harmonize these three dimensions. MASPO integrates a differentiable soft Gaussian gating to maximize gradient utility, a mass-adaptive limiter to balance exploration across the probability spectrum, and an asymmetric risk controller to align update magnitudes with signal confidence. Extensive evaluations demonstrate that MASPO serves as a robust, all-in-one RLVR solution, significantly outperforming baselines. Our code is available at: https://github.com/FlyTune/MASPO-RL.