Add new paper for Sec4.1 Task: Search Agent and Sec4.2 Task: Code Agent

[weiyifan1023]

Hi, thanks for maintaining this awesome list!
I would like to recommend adding the following recent RL-based Agent paper:

• Title: AutoTIR: Autonomous Tools Integrated Reasoning via Reinforcement Learning

• Intro: AutoTIR, a reinforcement learning framework that enables LLMs to autonomously decide whether and which tool (e.g. Search tool and Code tool) to invoke during the reasoning process, rather than following static tool-use strategies.

• Paper: https://arxiv.org/pdf/2507.21836

• Code: https://github.com/weiyifan1023/AutoTIR

• Base LLM: Qwen2.5-7B-Instruct

[xhyumiracle]

Thanks for raising this work and the quick intro!
One thing we'd like to be aligned with you before proceeding: Besides Seach task which I agree, I would recommend adding it under the Math task (rather than Code task), similar to ToRL, ZeroTIR etc.

Your thoughts?

[xhyumiracle]

I decide to proceed in order to include the work as early as possible.
We'll leave this issue open a longer time for your feedback.

Cheers.