← Back to all news
arXiv AI By Min-Seon Kim, Seungdong Yoa, Sanghyu Yoon, Ye Seul Sim, Suhee Yoon, Dongmin Kim, Jung-Eun Kim

Agents, Think Only When it Helps

Read the original on arXiv AI →

The paper investigates the role of thought supervision in distilling large language model agents. By independently perturbing teacher thoughts and actions during distillation, the authors find that removing or altering thoughts has little impact on student performance across various benchmarks and model scales. However, thought supervision becomes important when the action does not reveal the information contained in the thought, suggesting that including thoughts in the training data may be unnecessary for many cases.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv AI
Oct 26

Beyond On-Policy vs. Off-Policy: At What Level Should LLM Agent Distillation Follow the Student?

The paper proposes a hierarchical framework for distilling large language model agents, distinguishing between the policy that generates interaction histories and the policy that produces token prefixes for actions. It explores various combinations of teacher-generated and student-generated data across multiple models and environments, finding that pairing student-induced histories with teacher-generated action prefixes yields the strongest training signal, while fully on-policy data is only beneficial when mixed with other configurations. The study also shows that the specific pairing of teacher and student proportions at each level significantly impacts performance, and that strategically front-loading teacher-generated contexts and delaying on-policy contexts improves results.

By Ya Gao, Zhengtao Zou, Pekka Marttinen
llmsagentsefficiency
More like this →
arXiv AI
Sep 30

Distilling Agentic Systems: A Roadmap across Models, Artifacts, and Harnesses

The paper introduces Agent Distillation, a framework for transferring task‑solving knowledge from a teacher agent to a student agent. It categorizes where this knowledge is retained—within the model, as artifacts, through the execution harness, or across substrates—distinguishing transfer evidence from outcomes. An evaluation framework is proposed to link retention to causal contribution and practical utility, aiming to support reliable, maintainable, and safe agent development.

By Ziluowen Luo, Senzhang Wang, Chaozhuo Li, Jun Yin, Hao Yan, Ming Cheng, Chenxu Wang, Songyang Liu, Litian Zhang, Qiwei Ye, Zheng Liu, Philip S. Yu
agentsefficiency
More like this →
arXiv AI
Oct 5

Harness-Aware Distillation for Small Language Model Agents

The paper introduces Harness-Aware Distillation (HAD), a method for training smaller language model agents that preserves the surrounding harness—software managing context, tools, and feedback—while focusing distillation on the teacher’s contributions beyond the harness. HAD combines an action preference that contrasts teacher actions with and without harness information, and a validity check that filters out contradictory preference pairs. Experiments on long-horizon agent benchmarks show that HAD outperforms standard on‑policy distillation, reducing unproductive loops and improving error recovery without requiring task rewards or future information.

By Moonseok Choi, Taehong Moon, Giung Nam, Juho Lee
llmsagentsefficiencybenchmarks
More like this →
arXiv Machine Learning
Aug 4

Look Ahead Before You Distill: Future Trajectory Validation of Teacher Guidance for Agentic On-Policy Distillation

arXiv:2608. 01953v1 Announce Type: cross Abstract: On-policy distillation (OPD) provides teacher supervision on states visited by the student, reducing the distribution gap between training and inference.

By Chishui Chen, Yaoyou Fan, Te Sun, Yi Yang, Chenghao Sun, Delin Mao, Hongbo Qiao, Zuowei Zhang, Junxi Wang, Chenxing Sun, Yangen Hu, Lu Pan, Xuyang Liu, Linfeng Zhang
agentsefficiency
More like this →