Qinyu LuoBYTEDANCE SEED
Let’s talk
All research

ICLR 2025 Conference paper

Proactive Agent: Shifting LLM Agents from Reactive Responses to Active Assistance

Learning when an agent should offer help, using observations of user activity and feedback on proposed tasks.

Yaxi Lu, Shenzhi Yang, Cheng Qian, Guirong Chen, Qinyu Luo, Yesai Wu, Huadong Wang, Xin Cong, Zhong Zhang, Yankai Lin, Weiwen Liu, Yasheng Wang, Zhiyuan Liu, Fangming Liu, Maosong Sun

Fig. 3 — Original proactive agent framework.Original source ↗
Proactive Agent
Proactive Agent framework connecting environment events, task prediction, user feedback, and memory.

Fig. 3 — Original proactive agent framework.

01 / The question

Can an agent recognize a useful next step before the user explicitly asks for help?

02 / The approach

Proactive Agent

The work combines activity-based task prediction, human acceptance labels, a learned evaluator, and the ProactiveBench data pipeline. Fine-tuning studies examine how these signals help agents decide when and how to offer assistance.

Read the full paper

03 / Citation

Cite this work

Select the citation to copy it manually.

04 / Versions & records

Publication record

1 Google Scholar record

READ HERE · 27 PAGES

Full paper

Document

Loading document…

Scroll through every page. Select text to copy, or use search to find a passage.

More research

View all
EMNLP 2024

RepoAgent

An open-source agent that generates and maintains documentation with a view of the whole code repository.

arXiv 2026

One-Shot On-Policy Distillation

Investigating why on-policy distillation can recover much of full-data training’s gain from a single query.

GET IN TOUCH

Let’s talk.

For research, collaboration, or a conversation. Email me or connect on LinkedIn and X.

Open email app

You can also select and copy the address above.

LinkedInXGitHubGoogle Scholar