# Medical Diagnosis and Reasoning on MEDFOUND (OOD) Source: [Medical Diagnosis and Reasoning on MEDFOUND (OOD)]() Collection index: [llms.txt]() Data updated: 2026-04-14T11:22:35.692Z Evaluation of model performance on the MEDFOUND dataset focusing on Out-of-Distribution (OOD) medical diagnosis and reasoning capabilities, measured by DxCode Accuracy, T-Eval Accuracy, and TrustScore. 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 Diagnosis and Reasoning | | Dataset (including any stated split) | MEDFOUND (OOD) | | Metric | DxCode Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 13 of 13 | | 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 - [DxCode Accuracy](): Higher is better; unit: PERCENT. - [T-Eval Accuracy](): Higher is better; unit: PERCENT. - [TrustScore](): Higher is better; unit: PERCENT. ## Recorded results | Method | DxCode Accuracy | Result date | Report paper | | --- | --- | --- | --- | | CGCL | 29.83 | Not recorded | 69ddca6c7d0172140ec79edb | | CGCL | 29.28 | Not recorded | 69ddca6c7d0172140ec79edb | | GRPO | 29.01 | Not recorded | 69ddca6c7d0172140ec79edb | | SFT-CGCL(Stage3) | 28.45 | Not recorded | 69ddca6c7d0172140ec79edb | | GRPO | 28.17 | Not recorded | 69ddca6c7d0172140ec79edb | | DPO | 28.15 | Not recorded | 69ddca6c7d0172140ec79edb | | Trust-Think & Answer | 27.9 | Not recorded | 69ddca6c7d0172140ec79edb | | SFT-CGCL(Stage3) | 27.35 | Not recorded | 69ddca6c7d0172140ec79edb | | Trust-Think & Answer | 26.24 | Not recorded | 69ddca6c7d0172140ec79edb | | SFT-CoT | 25.1 | Not recorded | 69ddca6c7d0172140ec79edb | | SFT-GT | 24.38 | Not recorded | 69ddca6c7d0172140ec79edb | | Think & Answer | 23.12 | Not recorded | 69ddca6c7d0172140ec79edb | | Vanilla | 19.85 | Not recorded | 69ddca6c7d0172140ec79edb | ### CGCL DxCode Accuracy: 29.83. Result ID: 69de237047f34ce924c69555. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Ours","Backbone":"LLaMA3.2-3B-Instruct","Zero-shot":"true"}. This record does not assert independent reproduction. ### CGCL DxCode Accuracy: 29.28. Result ID: 69de237047f34ce924c69562. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"Ours"}. This record does not assert independent reproduction. ### GRPO DxCode Accuracy: 29.01. Result ID: 69de237047f34ce924c69561. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"RL"}. This record does not assert independent reproduction. ### SFT-CGCL(Stage3) DxCode Accuracy: 28.45. Result ID: 69de237047f34ce924c69553. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"SFT","Backbone":"LLaMA3.2-3B-Instruct","Zero-shot":"true"}. This record does not assert independent reproduction. ### GRPO DxCode Accuracy: 28.17. Result ID: 69de237047f34ce924c69554. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"RL","Backbone":"LLaMA3.2-3B-Instruct","Zero-shot":"true"}. This record does not assert independent reproduction. ### DPO DxCode Accuracy: 28.15. Result ID: 69de237047f34ce924c69560. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"RL"}. This record does not assert independent reproduction. ### Trust-Think & Answer DxCode Accuracy: 27.9. Result ID: 69de237047f34ce924c69552. [Reporting paper]() Recorded evaluation configuration: {"Category":"Prompt","Backbone":"LLaMA3.2-3B-Instruct","Zero-shot":"true"}. This record does not assert independent reproduction. ### SFT-CGCL(Stage3) DxCode Accuracy: 27.35. Result ID: 69de237047f34ce924c6955f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"SFT"}. This record does not assert independent reproduction. ### Trust-Think & Answer DxCode Accuracy: 26.24. Result ID: 69de237047f34ce924c6955c. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"Prompt"}. This record does not assert independent reproduction. ### SFT-CoT DxCode Accuracy: 25.1. Result ID: 69de237047f34ce924c6955e. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"SFT"}. This record does not assert independent reproduction. ### SFT-GT DxCode Accuracy: 24.38. Result ID: 69de237047f34ce924c6955d. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"SFT"}. This record does not assert independent reproduction. ### Think & Answer DxCode Accuracy: 23.12. Result ID: 69de237047f34ce924c6955b. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"Prompt"}. This record does not assert independent reproduction. ### Vanilla DxCode Accuracy: 19.85. Result ID: 69de237047f34ce924c6955a. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-3B-Instruct","Zero-shot":"true","Category":"Prompt"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.