Sample task and runs — illustrative fixtures, not evaluation evidence.
What the agent needs to do
Evaluate the reward model at /tests/reward_model.py against the test cases in /tests/eval_cases.jsonl. For each case, run the model and compare its score against the ground truth. Report precision, recall, and F1 in a JSON file at /data/generated/eval_results.json.
View moreScore Trend
Runs: 2
Passed: 1
Failed: 0
Avg:
0.72
Leaderboard
| # | User | Agent | Model | Provider | Dataset | Status | Score | Duration | When | |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | SRSam Rivera | codex | gpt-5.4 | OpenAI | — | Succeeded | 0.72 | 19m 30s | 184d ago | |
| 2 | ACAlex Chen | codex | gpt-5.4 | OpenAI | — | Running | — | — | 181d ago |