Building Math Agents with Multi-Turn Iterative Preference Learning
- URL: http://arxiv.org/abs/2409.02392v1
- Date: Wed, 4 Sep 2024 02:41:04 GMT
- Title: Building Math Agents with Multi-Turn Iterative Preference Learning
- Authors: Wei Xiong, Chengshuai Shi, Jiaming Shen, Aviv Rosenberg, Zhen Qin, Daniele Calandriello, Misha Khalman, Rishabh Joshi, Bilal Piot, Mohammad Saleh, Chi Jin, Tong Zhang, Tianqi Liu,
- Abstract summary: This paper studies the complementary direct preference learning approach to further improve model performance.
Existing direct preference learning algorithms are originally designed for the single-turn chat task.
We introduce a multi-turn direct preference learning framework, tailored for this context.
- Score: 56.71330214021884
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Recent studies have shown that large language models' (LLMs) mathematical problem-solving capabilities can be enhanced by integrating external tools, such as code interpreters, and employing multi-turn Chain-of-Thought (CoT) reasoning. While current methods focus on synthetic data generation and Supervised Fine-Tuning (SFT), this paper studies the complementary direct preference learning approach to further improve model performance. However, existing direct preference learning algorithms are originally designed for the single-turn chat task, and do not fully address the complexities of multi-turn reasoning and external tool integration required for tool-integrated mathematical reasoning tasks. To fill in this gap, we introduce a multi-turn direct preference learning framework, tailored for this context, that leverages feedback from code interpreters and optimizes trajectory-level preferences. This framework includes multi-turn DPO and multi-turn KTO as specific implementations. The effectiveness of our framework is validated through training of various language models using an augmented prompt set from the GSM8K and MATH datasets. Our results demonstrate substantial improvements: a supervised fine-tuned Gemma-1.1-it-7B model's performance increased from 77.5% to 83.9% on GSM8K and from 46.1% to 51.2% on MATH. Similarly, a Gemma-2-it-9B model improved from 84.1% to 86.3% on GSM8K and from 51.0% to 54.5% on MATH.
Related papers
- Kimi k1.5: Scaling Reinforcement Learning with LLMs [84.2229964736678]
We report on the training practice of Kimi k1.5, our latest multi-modal language model trained with reinforcement learning.
Long context scaling and improved policy optimization methods are key ingredients of our approach.
Our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities.
arXiv Detail & Related papers (2025-01-22T02:48:14Z) - Chain-of-Reasoning: Towards Unified Mathematical Reasoning in Large Language Models via a Multi-Paradigm Perspective [90.86370957353911]
Chain-of-Reasoning (CoR) is a novel unified framework that integrates multiple reasoning paradigms.
CoR generates multiple potential answers using different reasoning paradigms and synthesizes them into a coherent final solution.
Experimental results demonstrate that CoR-Math-7B significantly outperforms current SOTA models.
arXiv Detail & Related papers (2025-01-19T16:53:26Z) - Masked Thought: Simply Masking Partial Reasoning Steps Can Improve Mathematical Reasoning Learning of Language Models [102.72940700598055]
In reasoning tasks, even a minor error can cascade into inaccurate results.
We develop a method that avoids introducing external resources, relying instead on perturbations to the input.
Our training approach randomly masks certain tokens within the chain of thought, a technique we found to be particularly effective for reasoning tasks.
arXiv Detail & Related papers (2024-03-04T16:21:54Z) - MathGenie: Generating Synthetic Data with Question Back-translation for Enhancing Mathematical Reasoning of LLMs [38.127313175508746]
MathGenie is a novel method for generating diverse and reliable math problems from a small-scale problem-solution dataset.
Various pretrained models, ranging from 7B to 70B, are trained on the newly curated data to test the effectiveness of the proposed augmentation technique.
MathGenieLM-InternLM2 achieves an accuracy of 87.7% on GSM8K and 55.7% on MATH, securing the best overall score among open-source language models.
arXiv Detail & Related papers (2024-02-26T07:17:25Z) - Tool-Augmented Reward Modeling [58.381678612409]
We propose a tool-augmented preference modeling approach, named Themis, to address limitations by empowering RMs with access to external environments.
Our study delves into the integration of external tools into RMs, enabling them to interact with diverse external sources.
In human evaluations, RLHF trained with Themis attains an average win rate of 32% when compared to baselines.
arXiv Detail & Related papers (2023-10-02T09:47:40Z) - ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving [170.7899683843177]
ToRA is a series of Tool-integrated Reasoning Agents designed to solve challenging mathematical problems.
ToRA models significantly outperform open-source models on 10 mathematical reasoning datasets across all scales.
ToRA-Code-34B is the first open-source model that achieves an accuracy exceeding 50% on MATH.
arXiv Detail & Related papers (2023-09-29T17:59:38Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.