Continual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM Agents
Paper • 2610.33956 • Published • 39
CLE Lite — Rubrics
The evaluation rubrics and ground-truth references used to score CLE Lite.
This repository contains the evaluation rubrics (eval/t1.md to eval/t5.md) and ground-truth references (eval/gt/) for the twenty CLE Lite tasks. Access requires manual approval. To inspect the public assignments and workspaces first, see CLE Lite and CLE Lite Workspaces.
🗂️ The four CLE Lite repositories
| Repository | Contents | Access |
|---|---|---|
| CLE Lite | 100 task cards: assignments, expected inputs and outputs, input-file inventories | Public |
| CLE Lite Workspaces | Level instructions and the workspace files supplied to agents | Public |
| CLE Lite Rubrics (this repository) | Evaluation rubrics and ground-truth references | Approval required |
| CLE Lite Full | Complete task packages for running and scoring agents | Approval required |
📁 Layout
016-data-scientist/
└── eval/
├── t1.md … t5.md rubric for each level
└── gt/ ground-truth references
⚠️ Terms of use
Approved access to the rubrics is for research and evaluation work. Keep the rubrics and answers out of agent input and training data. CLE Lite Full is separately approval-gated.
📝 Citation
If you use these CLE Lite resources, please cite the CLE paper:
@misc{zhao2026continuallearningexambenchmarking,
title = {Continual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM Agents},
author = {Haoyu Zhao and Sitian Tu and Yuxuan Huang and Huichi Zhou
and Zhangyue Yin and Jiahao Huang and Jingyuan Xu and Tianmiao Shi
and Jin Shi and Nianxi Zhang and Yangmiao Zhang and Zhuolang Chan
and Huiyan Hu and Chenchao Sheng and Yaqing Luo and Yue Chen
and Huaxiu Yao and Meng Fang and Jun Wang},
year = {2026},
eprint = {2610.33956},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2610.33956}
}
- Downloads last month
- 51