# Scaling Agents via Continual Pre-training

Source: [Scaling Agents via Continual Pre-training](<https://www.sota2.com/research/paper/68fa7b671e209131cea8833d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-28T13:57:21.819Z

Authors: Liangcai Su, Zhen Zhang, Guangyu Li, Zhuo Chen, Chenxi Wang, Maojia Song, Xinyu Wang, Kuan Li, Jialong Wu, Xuanzhong Chen, Zile Qiao, Zhongwang Zhang, Huifeng Yin, Shihao Cai, Runnan Fang, Zhengwei Tao, Wenbiao Yin, Chenxiong Qian, Yong Jiang, Pengjun Xie, Fei Huang, Jingren Zhou

Published: 2025-09-16T17:57:19.000Z

## Abstract

Large language models (LLMs) have evolved into agentic systems capable of autonomous tool use and multi-step reasoning for complex problem-solving. However, post-training approaches building upon general-purpose foundation models consistently underperform in agentic tasks, particularly in open-source implementations. We identify the root cause: the absence of robust agentic foundation models forces models during post-training to simultaneously learn diverse agentic behaviors while aligning them to expert demonstrations, thereby creating fundamental optimization tensions. To this end, we are the first to propose incorporating Agentic Continual Pre-training (Agentic CPT) into the deep research agents training pipeline to build powerful agentic foundational models. Based on this approach, we develop a deep research agent model named AgentFounder. We evaluate our AgentFounder-30B on 10 benchmarks and achieve state-of-the-art performance while retains strong tool-use ability, notably 39.9% on BrowseComp-en, 43.3% on BrowseComp-zh, and 31.5% Pass@1 on HLE.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Long-horizon agentic tasks on HLE Full](<https://www.sota2.com/research/sota/long-horizon-agentic-tasks-on-hle-full.md>): Long-horizon agentic tasks; HLE Full; Pass@1: 31.5.

- [Long-horizon agentic tasks on BrowseComp Full](<https://www.sota2.com/research/sota/long-horizon-agentic-tasks-on-browsecomp-full.md>): Long-horizon agentic tasks; BrowseComp Full; Pass@1: 39.9.

- [Long-horizon agentic tasks on BrowseComp-ZH Our Settings](<https://www.sota2.com/research/sota/long-horizon-agentic-tasks-on-browsecomp-zh-our-settings.md>): Long-horizon agentic tasks; BrowseComp-ZH Our Settings; Pass@1: 43.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2509.13310>)

[PDF](<https://arxiv.org/pdf/2509.13310.pdf>)