# Medical Response Refinement on HealthBench 254 medical queries Source: [Medical Response Refinement on HealthBench 254 medical queries]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:37:39.962Z Evaluation of model performance on refining medical responses using the specific 254 queries subset of the HealthBench dataset, focusing on improvement metrics like base vs. refined scores and statistical significance. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Medical Response Refinement | | Dataset (including any stated split) | HealthBench 254 medical queries | | Metric | Base Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 4 of 4 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Base Score](): Higher is better; unit: PERCENT. - [Refined Score](): Higher is better; unit: PERCENT. - [Absolute Improvement (Δ)](): Higher is better; unit: SCALAR. - [Relative Improvement (%)](): Higher is better; unit: PERCENT. - [p-value](): Lower is better; unit: SCALAR. ## Recorded results | Method | Base Score | Result date | Report paper | | --- | --- | --- | --- | | Self-Critique | 59 | Not recorded | 6971af37e25ad2fa9d1edbc9 | | GPT-4o Rubrics | 59 | Not recorded | 6971af37e25ad2fa9d1edbc9 | | Our Rubrics | 59 | Not recorded | 6971af37e25ad2fa9d1edbc9 | | Reference Rubrics | 58.9 | Not recorded | 6971af37e25ad2fa9d1edbc9 | ### Self-Critique Base Score: 59. Result ID: 6996763e5698665ec4bdbccf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Guidance Source":"Intrinsic self-correction (No external rubric)","Role":"Lower-bound control","Base Response Model":"Llama-3.1-8B-Instant","Evaluator Model":"Llama-3.3-70B","Editor Model":"Llama-3.1-8B","Protocol":"Critique-then-Refine"}. This record does not assert independent reproduction. ### GPT-4o Rubrics Base Score: 59. Result ID: 6996763e5698665ec4bdbcd0. [Reporting paper]() Recorded evaluation configuration: {"Guidance Source":"GPT-4o generated rubrics","Base Response Model":"Llama-3.1-8B-Instant","Evaluator Model":"Llama-3.3-70B","Editor Model":"Llama-3.1-8B","Protocol":"Critique-then-Refine"}. This record does not assert independent reproduction. ### Our Rubrics Base Score: 59. Result ID: 6996763e5698665ec4bdbcd1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Guidance Source":"Proposed instance-specific rubrics","Base Response Model":"Llama-3.1-8B-Instant","Evaluator Model":"Llama-3.3-70B","Editor Model":"Llama-3.1-8B","Protocol":"Critique-then-Refine"}. This record does not assert independent reproduction. ### Reference Rubrics Base Score: 58.9. Result ID: 6996763e5698665ec4bdbcce. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Guidance Source":"Expert physician-authored rubrics","Role":"Oracle Upper Bound","Base Response Model":"Llama-3.1-8B-Instant","Evaluator Model":"Llama-3.3-70B","Editor Model":"Llama-3.1-8B","Protocol":"Critique-then-Refine"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.