| CORRECT: Condensed Error Recognition via Knowledge Transfer in Multi-agent Systems |
| Copyright 2026 The CORRECT Authors |
|
|
| This product includes software and adapted code from the following |
| third-party projects. Their original licenses apply to their respective |
| portions. |
|
|
| ------------------------------------------------------------------------------- |
| 1. Agents_Failure_Attribution (Who&When) |
| https: |
|
|
| The inference and error-template-generation scripts under `src/` are |
| adapted from this repository. We retain the structure of the upstream |
| pipeline (trajectory parsing, similarity-based template retrieval, |
| all-at-once / step-by-step / binary-search detection methods) and |
| extend it with CORRECT's schema-guided detection. |
|
|
| Cite the upstream Who&When work: |
| Zhang, S., Yin, M., Zhang, J., Liu, J., Han, Z., Zhang, J., Li, B., |
| Wang, C., Wang, H., Chen, Y., and Wu, Q. (2025). |
| "Which Agent Causes Task Failures and When? On Automated Failure |
| Attribution of LLM Multi-Agent Systems." ICML 2025. |
|
|
| ------------------------------------------------------------------------------- |
| 2. Magentic-One |
| https: |
| Microsoft Research |
|
|
| The multi-agent trajectories distributed as part of the CORRECT-Error |
| benchmark were generated by running Magentic-One on standard benchmark |
| tasks (GAIA, HotpotQA, Musique, WikiMQA, ARC, Math500, MMLU-Pro). |
| Magentic-One is released under the MIT License. |
|
|
| Cite Magentic-One: |
| Fourney, A., Bansal, G., Mozannar, H., Tan, C., Salinas, E., Niedtner, |
| F., Proebsting, G., Bassman, G., Gerrits, J., Alber, J., et al. (2024). |
| "Magentic-One: A Generalist Multi-Agent System for Solving Complex |
| Tasks." arXiv:2411.04468. |
|
|
| ------------------------------------------------------------------------------- |
|
|
