Discovery and Reinforcement of Tool-Integrated Reasoning Chains via Rollout Trees
This addresses the challenge of augmenting LLMs with computational tools for complex reasoning tasks, representing a novel method for a known bottleneck rather than a foundational advancement.
The paper tackles the problem of integrating tool-use into long Chain-of-Thought reasoning for Large Language Models, which is underexplored due to data scarcity and integration challenges, and introduces DART, a reinforcement learning framework that enables spontaneous tool-use without human annotation, significantly outperforming existing methods on benchmarks like AIME and GPQA-Diamond.
Tool-Integrated Reasoning has emerged as a key paradigm to augment Large Language Models (LLMs) with computational capabilities, yet integrating tool-use into long Chain-of-Thought (long CoT) remains underexplored, largely due to the scarcity of training data and the challenge of integrating tool-use without compromising the model's intrinsic long-chain reasoning. In this paper, we introduce DART (Discovery And Reinforcement of Tool-Integrated Reasoning Chains via Rollout Trees), a reinforcement learning framework that enables spontaneous tool-use during long CoT reasoning without human annotation. DART operates by constructing dynamic rollout trees during training to discover valid tool-use opportunities, branching out at promising positions to explore diverse tool-integrated trajectories. Subsequently, a tree-based process advantage estimation identifies and credits specific sub-trajectories where tool invocation positively contributes to the solution, effectively reinforcing these beneficial behaviors. Extensive experiments on challenging benchmarks like AIME and GPQA-Diamond demonstrate that DART significantly outperforms existing methods, successfully harmonizing tool execution with long CoT reasoning.