AgentDistill: Training-Free Agent Distillation with Generalizable MCP Boxes
2025
Read AgentDistill: Training-Free Agent Distillation with Generalizable MCP Boxes (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-14728.txt.
Description
Jiahao Qiu and co-authors distill big agents into small ones without training in a 2025 paper. Agent skills like planning, memory, and tool use resist the output matching tricks used to shrink plain models. Past agent distill work replays whole teacher trails, which leaves students stiff in new settings.
Their frame, AgentDistill, hands students ready made task blocks called MCP boxes, built by the teacher on its own. Students reuse those blocks to solve fresh tasks with slight watch. On biomedical and math tests, small model students matched big model systems like OctoTools. The route makes cheap agents that still plan and act.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-14728