ByteDance-Seed/EdgeBench
Overview EdgeBench is a benchmark of 134 real-world tasks for evaluating how autonomous AI agents learn from real-world environments. Instead of measuring one-shot performance, EdgeBench places agents in executable task environments with realistic, multi-level feedback and lets them iterate for 12+ hours per task — tracking the full trajectory of improvement, not just the final score. We publicly release 51 tasks… See the full description on the dataset page: https://huggingface.co/datasets/ByteDance-Seed/EdgeBench.
847.5k
1{2 "task_id": "k12_math_recommendation",3 "name": "K12 Math Recommendation",4 "category": "Professional Knowledge Work",5 "base_image": "python",6 "platform": "linux/amd64",7 "internet": false,8 "cwd": "/home/workspace",9 "submit_paths": [10 "train.py",11 "infer.py",12 "models",13 "submission.json",14 "requirements.txt",15 "README.md"16 ],17 "submit_exclude": [18 ".git",19 "__pycache__",20 "*.pyc",21 "node_modules",22 "bin",23 "obj"24 ],25 "work": {26 "image_tag": "fecdcfb17904",27 "specs_dir": null,28 "agent_query": "Read the complete task instructions in `/home/workspace/task_instruction.md`, and complete the task according to those requirements. The final deliverables must be written to the file paths specified in `task_instruction.md`.\n"29 },30 "judge": {31 "image_tag": "1faabfecdb8e",32 "eval_cmd": "cd /home/workspace && python3 scoring/score.py",33 "eval_timeout": 900,34 "parser": "structured_json",35 "score_direction": "maximize",36 "selection": "score_first"37 }38}39