Salvato in:
Dettagli Bibliografici
Autori principali: Hou, Zhenyu, Lv, Xin, Lu, Rui, Zhang, Jiajie, Li, Yujiang, Yao, Zijun, Li, Juanzi, Tang, Jie, Dong, Yuxiao
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2501.11651
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909648036036608
author Hou, Zhenyu
Lv, Xin
Lu, Rui
Zhang, Jiajie
Li, Yujiang
Yao, Zijun
Li, Juanzi
Tang, Jie
Dong, Yuxiao
author_facet Hou, Zhenyu
Lv, Xin
Lu, Rui
Zhang, Jiajie
Li, Yujiang
Yao, Zijun
Li, Juanzi
Tang, Jie
Dong, Yuxiao
contents Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks. However, existing approaches mainly rely on imitation learning and struggle to achieve effective test-time scaling. While reinforcement learning (RL) holds promise for enabling self-exploration, recent attempts yield modest improvements in complex reasoning. In this paper, we present T1 to scale RL by encouraging exploration and understand inference scaling. We first initialize the LLM using synthesized chain-of-thought data that integrates trial-and-error and self-verification. To scale RL training, we promote increased sampling diversity through oversampling. We demonstrate that T1 with open LLMs as its base exhibits inference scaling behavior and achieves superior performance on challenging math reasoning benchmarks. More importantly, we present a simple strategy to examine inference scaling, where increased inference budgets directly lead to T1's better performance without any additional verification.
format Preprint
id arxiv_https___arxiv_org_abs_2501_11651
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle T1: Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling
Hou, Zhenyu
Lv, Xin
Lu, Rui
Zhang, Jiajie
Li, Yujiang
Yao, Zijun
Li, Juanzi
Tang, Jie
Dong, Yuxiao
Machine Learning
Computation and Language
Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks. However, existing approaches mainly rely on imitation learning and struggle to achieve effective test-time scaling. While reinforcement learning (RL) holds promise for enabling self-exploration, recent attempts yield modest improvements in complex reasoning. In this paper, we present T1 to scale RL by encouraging exploration and understand inference scaling. We first initialize the LLM using synthesized chain-of-thought data that integrates trial-and-error and self-verification. To scale RL training, we promote increased sampling diversity through oversampling. We demonstrate that T1 with open LLMs as its base exhibits inference scaling behavior and achieves superior performance on challenging math reasoning benchmarks. More importantly, we present a simple strategy to examine inference scaling, where increased inference budgets directly lead to T1's better performance without any additional verification.
title T1: Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2501.11651