Agentic-R1: Distilled Dual-Strategy Reasoning

TL;DR

Agentic-R1 employs a dual-strategy reasoning framework, combining multiple reasoning approaches through distillation to enhance accuracy and robustness in complex tasks, outperforming traditional single-strategy models.

Contribution

We introduce DualDistill, a novel fine-tuning method that distills multiple reasoning strategies into a single model capable of dynamic strategy selection.

Findings

Improved accuracy on diverse reasoning benchmarks.

Effective combination of text-based and tool-based reasoning.

Demonstrated robustness across computation and abstract tasks.

Abstract

Current long chain-of-thought (long-CoT) models excel at mathematical reasoning but rely on slow and error-prone natural language traces. Tool-augmented agents address arithmetic via code execution, but often falter on complex logical tasks. We introduce a fine-tuning framework, DualDistill, that distills complementary reasoning strategies from multiple teachers into a unified student model. Using this approach, we train Agentic-R1, which dynamically selects the optimal strategy for each query, invoking tools for arithmetic and algorithmic problems, and using text-based reasoning for abstract ones. Our method improves accuracy across a range of tasks, including both computation-intensive and standard benchmarks, demonstrating the effectiveness of multi-strategy distillation in achieving robust and efficient reasoning. Our project is available at https://github.com/StigLidu/DualDistill