Saved in:
Bibliographic Details
Main Authors: Li, Yu, Yi, Mingyang, Li, Xiuyu, Fan, Ju, Jiang, Fuxin, Chen, Binbin, Li, Peng, Song, Jie, Zhang, Tieying
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.00994
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910268256157696
author Li, Yu
Yi, Mingyang
Li, Xiuyu
Fan, Ju
Jiang, Fuxin
Chen, Binbin
Li, Peng
Song, Jie
Zhang, Tieying
author_facet Li, Yu
Yi, Mingyang
Li, Xiuyu
Fan, Ju
Jiang, Fuxin
Chen, Binbin
Li, Peng
Song, Jie
Zhang, Tieying
contents Agentic Reinforcement Learning (ARL) trains large language models to interleave reasoning with external tool execution to solve complex tasks. Most existing ARL methods train a single set of parameters to support both reasoning and tool-use behaviors, implicitly assuming that joint training leads to improved overall agent performance. Despite its widespread adoption, this assumption has rarely been examined empirically. In this paper, we systematically examine this assumption by introducing Capability Effect Attribution (CEA), which provides quantitative evidence of interference between reasoning and tool-use behaviors. Through an in-depth analysis, we show that these two capabilities often induce misaligned gradient directions, leading to training interference that undermines the effectiveness of joint optimization and challenges the prevailing ARL paradigm. To address this issue, we propose Disentangled Action--Reasoning Tuning (DART), a simple and efficient framework that explicitly decouples parameter updates for reasoning and tool use via separate low-rank adaptation modules. With this simple change alone, DART outperforms all joint-optimization baselines and approaches the 2-Agent upper bound across thirteen benchmarks on retrieval-augmented QA and NL2SQL, further supporting our finding of capability interference under shared optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2602_00994
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Reasoning and Tool-use Compete in Agentic RL:From Quantifying Interference to Disentangled Tuning
Li, Yu
Yi, Mingyang
Li, Xiuyu
Fan, Ju
Jiang, Fuxin
Chen, Binbin
Li, Peng
Song, Jie
Zhang, Tieying
Artificial Intelligence
Agentic Reinforcement Learning (ARL) trains large language models to interleave reasoning with external tool execution to solve complex tasks. Most existing ARL methods train a single set of parameters to support both reasoning and tool-use behaviors, implicitly assuming that joint training leads to improved overall agent performance. Despite its widespread adoption, this assumption has rarely been examined empirically. In this paper, we systematically examine this assumption by introducing Capability Effect Attribution (CEA), which provides quantitative evidence of interference between reasoning and tool-use behaviors. Through an in-depth analysis, we show that these two capabilities often induce misaligned gradient directions, leading to training interference that undermines the effectiveness of joint optimization and challenges the prevailing ARL paradigm. To address this issue, we propose Disentangled Action--Reasoning Tuning (DART), a simple and efficient framework that explicitly decouples parameter updates for reasoning and tool use via separate low-rank adaptation modules. With this simple change alone, DART outperforms all joint-optimization baselines and approaches the 2-Agent upper bound across thirteen benchmarks on retrieval-augmented QA and NL2SQL, further supporting our finding of capability interference under shared optimization.
title Reasoning and Tool-use Compete in Agentic RL:From Quantifying Interference to Disentangled Tuning
topic Artificial Intelligence
url https://arxiv.org/abs/2602.00994