Evaluation Criteria Generation on HealthBench
90CoverageQworld ret.
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qworld ret.Backbone=GPT-4.1, Retrieval Augmented=true2026.03 | 90 | 82 | 84 | 83 | |
| QworldBackbone=GPT-4.12026.03 | 89 | 79 | 83 | 85 | |
| EvalAgentBackbone=GPT-4.12026.03 | 83 | 50 | 40 | 65 | |
| OpenRubricsBackbone=GPT-4.12026.03 | 54 | 37 | 36 | 49 | |
| RocketEvalBackbone=GPT-4.12026.03 | 53 | 26 | 42 | 83 | |
| TICKBackbone=GPT-4.12026.03 | 46 | 24 | 29 | 79 |