# Medical Visual Question Answering and Reasoning on CheXbench Source: [Medical Visual Question Answering and Reasoning on CheXbench]() Collection index: [llms.txt]() Data updated: 2026-04-16T07:03:07.371Z Evaluation of model performance on the CheXbench dataset for Medical Visual Question Answering and Reasoning, reporting specialized accuracies including Rad-Restruct, SLAKE, and OpenI Reasoning. 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 Visual Question Answering and Reasoning | | Dataset (including any stated split) | CheXbench | | Metric | Rad-Restruct Acc | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 6 of 6 | | 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 - [Rad-Restruct Acc](): Higher is better; unit: PERCENT. - [SLAKE Acc](): Higher is better; unit: PERCENT. - [OpenI Reasoning Acc](): Higher is better; unit: PERCENT. - [Overall Acc](): Higher is better; unit: PERCENT. ## Recorded results | Method | Rad-Restruct Acc | Result date | Report paper | | --- | --- | --- | --- | | RadAgents | 76.5 | Not recorded | 68fa7b971e209131cea89585 | | Qwen3-VL w/ Workflow | 72.2 | Not recorded | 68fa7b971e209131cea89585 | | RadAgents wo/ V-RAG | 71.3 | Not recorded | 68fa7b971e209131cea89585 | | Qwen3-VL w/ ReAct | 70.4 | Not recorded | 68fa7b971e209131cea89585 | | CheXagent | 57.1 | Not recorded | 68fa7b971e209131cea89585 | | GPT-4o | 53.9 | Not recorded | 68fa7b971e209131cea89585 | ### RadAgents Rad-Restruct Acc: 76.5. Result ID: 69e0889847f34ce924c94976. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model / Backbone":"RadAgents","V-RAG":"true (k=3)"}. This record does not assert independent reproduction. ### Qwen3-VL w/ Workflow Rad-Restruct Acc: 72.2. Result ID: 69e0889847f34ce924c94974. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model / Backbone":"Qwen3-VL","Agent Workflow Strategy":"Workflow templates"}. This record does not assert independent reproduction. ### RadAgents wo/ V-RAG Rad-Restruct Acc: 71.3. Result ID: 69e0889847f34ce924c94975. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model / Backbone":"RadAgents","V-RAG":"false"}. This record does not assert independent reproduction. ### Qwen3-VL w/ ReAct Rad-Restruct Acc: 70.4. Result ID: 69e0889847f34ce924c94973. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model / Backbone":"Qwen3-VL","Agent Workflow Strategy":"ReAct"}. This record does not assert independent reproduction. ### CheXagent Rad-Restruct Acc: 57.1. Result ID: 69e0889847f34ce924c94971. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model / Backbone":"CheXagent"}. This record does not assert independent reproduction. ### GPT-4o Rad-Restruct Acc: 53.9. Result ID: 69e0889847f34ce924c94972. [Reporting paper]() Recorded evaluation configuration: {"Model / Backbone":"GPT-4o"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.