# Visual Spatial Reasoning on VSR ZOOM-Hard Source: [Visual Spatial Reasoning on VSR ZOOM-Hard]() Collection index: [llms.txt]() Data updated: 2026-05-28T08:21:47.888Z Evaluation of model performance on the ZOOM-Hard subset of the VSR dataset for Visual Spatial Reasoning, measured by GPT Accuracy and Incremental Accuracy. 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 | Visual Spatial Reasoning | | Dataset (including any stated split) | VSR ZOOM-Hard | | Metric | GPT Accuracy | | 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 - [GPT Accuracy](): Higher is better; unit: PERCENT. - [Incremental Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | GPT Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Mags-RL | 55.98 | Not recorded | 6a17cbf6dbd7150506f8119e | | Zoom-Refine | 52.72 | Not recorded | 6a17cbf6dbd7150506f8119e | | Chain-of-Thought | 46.74 | Not recorded | 6a17cbf6dbd7150506f8119e | | Direct Query | 45.65 | Not recorded | 6a17cbf6dbd7150506f8119e | | One-shot ICL | 45.11 | Not recorded | 6a17cbf6dbd7150506f8119e | | GRIT | 45.11 | Not recorded | 6a17cbf6dbd7150506f8119e | ### Mags-RL GPT Accuracy: 55.98. Result ID: 6a17faeafbc373592c11a954. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Zoom-Refine GPT Accuracy: 52.72. Result ID: 6a17faeafbc373592c11a952. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Chain-of-Thought GPT Accuracy: 46.74. Result ID: 6a17faeafbc373592c11a950. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Direct Query GPT Accuracy: 45.65. Result ID: 6a17faeafbc373592c11a94f. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### One-shot ICL GPT Accuracy: 45.11. Result ID: 6a17faeafbc373592c11a951. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"ICL shots":"1"}. This record does not assert independent reproduction. ### GRIT GPT Accuracy: 45.11. Result ID: 6a17faeafbc373592c11a953. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.