Reward Aggregation on HealthBench 500
65.7ScoreRuscaRL + GEAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RuscaRL + GEARPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 65.7 | 7.5 | |
| RuscaRL + GEARPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 63.8 | 6.9 | |
| GEARPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Aggregation-only2026.06 | 62.9 | 6.2 | |
| GEARPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Aggregation-only2026.06 | 62.5 | 8.4 | |
| RuscaRLPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 58.2 | — | |
| RuscaRLPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 56.9 | — | |
| Flat aggregationPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Aggregation-only2026.06 | 56.7 | — | |
| RaR-Explicit + GEARPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 55 | 5.5 | |
| Hard gatingPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Aggregation-only2026.06 | 54.6 | — | |
| Flat aggregationPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Aggregation-only2026.06 | 54.1 | — | |
| RaR-Explicit + GEARPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 54.1 | 5.7 | |
| Hard gatingPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Aggregation-only2026.06 | 53 | — | |
| RaR-ExplicitPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 49.5 | — | |
| RaR-ExplicitPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 48.4 | — | |
| MeRF + GEARPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 45.6 | 5.4 | |
| MeRF + GEARPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 44.7 | 5.2 | |
| MeRFPolicy Model=Qwen2.5-7B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 40.2 | — | |
| MeRFPolicy Model=Llama-3.1-8B-Instruct, Pipeline Category=Plug-in replacement in rubric-guided RL pipelines2026.06 | 39.5 | — |