Continual-Learning-Exam/CLE-Lite-Workspaces

CLE Lite — Workspaces

The level instructions and workspace files of the twenty CLE Lite tasks, as supplied to agents.

Public agent-facing materials for the twenty CLE Lite tasks. Each task directory contains five level instructions (test/t1.md to test/t5.md), the workspace files that arrive at each level (workspace/t1/ to workspace/t5/), and source notices (NOTICE.md, PROVENANCE.md). Evaluation rubrics, reference answers, setup scripts, and the evaluation harness are excluded.

🗂️ 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 (this repository)	Level instructions and the workspace files supplied to agents	Public
CLE Lite Rubrics	Evaluation rubrics and ground-truth references	Approval required
CLE Lite Full	Complete task packages for running and scoring agents	Approval required

Start with CLE Lite to browse all 100 task-level cards.

📁 Layout

016-data-scientist/
├── test/
│   └── t1.md … t5.md        level instructions
├── workspace/
│   └── t1/ … t5/            files that arrive at each level
├── NOTICE.md                third-party sources and terms
└── PROVENANCE.md            where each file came from

⚠️ Terms of use

The workspace files are source data, not a hosted service. Downloading this repository alone does not reproduce the benchmark environment. Please do not include these tasks in training corpora. Third-party sources and terms are listed in each task's NOTICE.md.

📝 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}
}

Download BibTeX · Evaluation reporting details