Search papers, labs, and topics across Lattice.
This paper analyzes reinforcement learning from verifiable rewards (RLVR) through a discriminative lens, revealing that standard RLVR updates can be diluted by shared, high-frequency token patterns. To address this, they introduce DelTA, a method that estimates token coefficients to amplify side-specific token-gradient directions and downweight shared ones. Experiments across mathematical benchmarks, code generation, and out-of-domain settings demonstrate that DelTA significantly improves performance compared to strong baselines.
Standard RLVR's token-level updates are often dominated by irrelevant tokens, but DelTA's discriminative reweighting unlocks significant performance gains in reasoning tasks.
Reinforcement learning from verifiable rewards (RLVR) has emerged as a central technique for improving the reasoning capabilities of large language models. Despite its effectiveness, how response-level rewards translate into token-level probability changes remains poorly understood. We introduce a discriminator view of RLVR updates, showing that the policy-gradient update direction implicitly acts as a linear discriminator over token-gradient vectors and thereby determines which token probabilities are increased or decreased during learning. Under standard sequence-level RLVR, this discriminator is constructed from positive- and negative-side centroids formed by advantage-weighted averaging of token-gradient vectors. However, such centroid construction can be dominated by shared high-frequency patterns, such as formatting tokens, diluting sparse yet discriminative directions that better distinguish high-reward responses from low-reward ones. To address this limitation, we propose DelTA, a discriminative token credit assignment method that estimates token coefficients to amplify side-specific token-gradient directions and downweight shared or weakly discriminative ones. These coefficients reweight a self-normalized RLVR surrogate, making the effective side-wise centroids more contrastive and thereby reshaping the RLVR update direction. On seven mathematical benchmarks, DelTA outperforms the strongest same-scale baselines by 3.26 and 2.62 average points on Qwen3-8B-Base and Qwen3-14B-Base, respectively. Additional results on code generation, a different backbone, and out-of-domain evaluations further demonstrate the generalization ability of DelTA.
