Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Rojas, Kevin, Lin, Jiahe, Rasul, Kashif, Schneider, Anderson, Nevmyvaka, Yuriy, Tao, Molei, Deng, Wei
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915789770063872
author Rojas, Kevin
Lin, Jiahe
Rasul, Kashif
Schneider, Anderson
Nevmyvaka, Yuriy
Tao, Molei
Deng, Wei
author_facet Rojas, Kevin
Lin, Jiahe
Rasul, Kashif
Schneider, Anderson
Nevmyvaka, Yuriy
Tao, Molei
Deng, Wei
contents Diffusion language models (DLMs) enable parallel, order-agnostic generation with iterative refinement, offering a flexible alternative to autoregressive large language models (LLMs). However, adapting reinforcement learning (RL) fine-tuning to DLMs remains an open challenge because of the intractable likelihood. Pioneering work such as diffu-GRPO estimated token-level likelihoods via one-step unmasking. While computationally efficient, this approach is severely biased. A more principled foundation lies in sequence-level likelihoods, where the evidence lower bound (ELBO) serves as a surrogate. Yet, despite this clean mathematical connection, ELBO-based methods have seen limited adoption due to the prohibitive cost of likelihood evaluation. In this work, we revisit ELBO estimation and disentangle its sources of variance. This decomposition motivates reducing variance through fast, deterministic integral approximations along a few pivotal dimensions. Building on this insight, we introduce Group Diffusion Policy Optimization (GDPO), a new RL algorithm tailored for DLMs. GDPO leverages simple yet effective Semi-deterministic Monte Carlo schemes to mitigate the variance explosion of ELBO estimators under vanilla double Monte Carlo sampling, yielding a provably lower-variance estimator under tight evaluation budgets. Empirically, GDPO achieves consistent gains over pretrained checkpoints and outperforms diffu-GRPO, one of the state-of-the-art baselines, on the majority of math, reasoning, and coding benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2510_08554
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
Rojas, Kevin
Lin, Jiahe
Rasul, Kashif
Schneider, Anderson
Nevmyvaka, Yuriy
Tao, Molei
Deng, Wei
Machine Learning
Diffusion language models (DLMs) enable parallel, order-agnostic generation with iterative refinement, offering a flexible alternative to autoregressive large language models (LLMs). However, adapting reinforcement learning (RL) fine-tuning to DLMs remains an open challenge because of the intractable likelihood. Pioneering work such as diffu-GRPO estimated token-level likelihoods via one-step unmasking. While computationally efficient, this approach is severely biased. A more principled foundation lies in sequence-level likelihoods, where the evidence lower bound (ELBO) serves as a surrogate. Yet, despite this clean mathematical connection, ELBO-based methods have seen limited adoption due to the prohibitive cost of likelihood evaluation. In this work, we revisit ELBO estimation and disentangle its sources of variance. This decomposition motivates reducing variance through fast, deterministic integral approximations along a few pivotal dimensions. Building on this insight, we introduce Group Diffusion Policy Optimization (GDPO), a new RL algorithm tailored for DLMs. GDPO leverages simple yet effective Semi-deterministic Monte Carlo schemes to mitigate the variance explosion of ELBO estimators under vanilla double Monte Carlo sampling, yielding a provably lower-variance estimator under tight evaluation budgets. Empirically, GDPO achieves consistent gains over pretrained checkpoints and outperforms diffu-GRPO, one of the state-of-the-art baselines, on the majority of math, reasoning, and coding benchmarks.
title Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
topic Machine Learning
url https://arxiv.org/abs/2510.08554