Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models

TL;DR

This paper introduces DCoLT, a novel reasoning framework for diffusion language models that enables bidirectional, non-linear reasoning, significantly improving performance on math and code generation tasks.

Contribution

The paper presents DCoLT, a new diffusion-based reasoning method that optimizes entire reasoning trajectories using reinforcement learning, differing from traditional linear Chain-of-Thought approaches.

Findings

DCoLT-reinforced DLMs outperform other models on multiple tasks.

LLaDA with DCoLT improves reasoning accuracy by up to 19.5%.

The approach effectively leverages diffusion models for complex reasoning tasks.

Abstract

We introduce the Diffusion Chain of Lateral Thought (DCoLT), a reasoning framework for diffusion language models. DCoLT treats each intermediate step in the reverse diffusion process as a latent "thinking" action and optimizes the entire reasoning trajectory to maximize the reward on the correctness of the final answer with outcome-based Reinforcement Learning (RL). Unlike traditional Chain-of-Thought (CoT) methods that follow a causal, linear thinking process, DCoLT allows bidirectional, non-linear reasoning with no strict rule on grammatical correctness amid its intermediate steps of thought. We implement DCoLT on two representative Diffusion Language Models (DLMs). First, we choose SEDD as a representative continuous-time discrete diffusion model, where its concrete score derives a probabilistic policy to maximize the RL reward over the entire sequence of intermediate diffusion steps. We further consider the discrete-time masked diffusion language model -- LLaDA, and find that the order to predict and unmask tokens plays an essential role to optimize its RL action resulting from the ranking-based Unmasking Policy Module (UPM) defined by the Plackett-Luce model. Experiments on both math and code generation tasks show that using only public data and 16 H800 GPUs, DCoLT-reinforced DLMs outperform other DLMs trained by SFT or RL or even both. Notably, DCoLT-reinforced LLaDA boosts its reasoning accuracy by +9.8%, +5.7%, +11.4%, +19.5% on GSM8K, MATH, MBPP, and HumanEval.