# Geometric Question Answering on GEOQA (test) Source: [Geometric Question Answering on GEOQA (test)]() Collection index: [llms.txt]() Data updated: 2026-06-15T07:22:31.754Z Evaluation of model performance on the test split of the GEOQA dataset for Geometric Question Answering, measured by success rates across geometric primitives (Points, Lines, Circles, Polygons). 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 | Geometric Question Answering | | Dataset (including any stated split) | GEOQA (test) | | Metric | Total Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 34 of 34 | | 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 - [Total Accuracy](): Higher is better; unit: PERCENT. - [Accuracy (Points)](): Higher is better; unit: PERCENT. - [Accuracy (Lines)](): Higher is better; unit: PERCENT. - [Accuracy (Circles)](): Higher is better; unit: PERCENT. - [Accuracy (Polygons)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Total Accuracy | Result date | Report paper | | --- | --- | --- | --- | | VeriGeo-beam5 | 82.74 | Not recorded | 6a2f86e0d33dfb23e783be19 | | VeriGeo-greedy | 79.31 | Not recorded | 6a2f86e0d33dfb23e783be19 | | TR-CoT-Qwen2.5-VL-7B | 79.2 | Not recorded | 6a2f86e0d33dfb23e783be19 | | GeoGen-SFT-7B | 78 | Not recorded | 6a2f86e0d33dfb23e783be19 | | TR-CoT-InternVL2.5-8B | 76.7 | Not recorded | 6a2f86e0d33dfb23e783be19 | | TR-CoT-8B | 75.9 | Not recorded | 6a2f86e0d33dfb23e783be19 | | GeoGen-SFT-3B | 75.6 | Not recorded | 6a2f86e0d33dfb23e783be19 | | Qwen2.5-VL-7B-Instruct | 72.07 | Not recorded | 6a2f86e0d33dfb23e783be19 | | MAVIS-7B | 68.3 | Not recorded | 6a2f86e0d33dfb23e783be19 | | G-LLaVA-13B | 67 | Not recorded | 6a2f86e0d33dfb23e783be19 | | G-LLaVA-7B | 62.8 | Not recorded | 6a2f86e0d33dfb23e783be19 | | GPT-4o | 61.4 | Not recorded | 6a2f86e0d33dfb23e783be19 | | InternVL2.5-8B | 59 | Not recorded | 6a2f86e0d33dfb23e783be19 | | InternVL2-8B | 56.5 | Not recorded | 6a2f86e0d33dfb23e783be19 | | Qwen2-VL-7B/8B | 55.7 | Not recorded | 6a2f86e0d33dfb23e783be19 | | FGRPO | 53.93 | Not recorded | 6a210658c10484bac23efe5c | | FedProx | 52.93 | Not recorded | 6a210658c10484bac23efe5c | | SCAFFOLD | 51.95 | Not recorded | 6a210658c10484bac23efe5c | | FedAvg | 51.4 | Not recorded | 6a210658c10484bac23efe5c | | FGRPO | 51.36 | Not recorded | 6a210658c10484bac23efe5c | | FedProx | 48.95 | Not recorded | 6a210658c10484bac23efe5c | | FedAvg | 48.87 | Not recorded | 6a210658c10484bac23efe5c | | SCAFFOLD | 48.42 | Not recorded | 6a210658c10484bac23efe5c | | Math-LLaVA-13B | 47.8 | Not recorded | 6a2f86e0d33dfb23e783be19 | | GPT-4V | 43.4 | Not recorded | 6a2f86e0d33dfb23e783be19 | | FGRPO | 40.76 | Not recorded | 6a210658c10484bac23efe5c | | FedProx | 39.28 | Not recorded | 6a210658c10484bac23efe5c | | SCAFFOLD | 38.08 | Not recorded | 6a210658c10484bac23efe5c | | FedAvg | 35.59 | Not recorded | 6a210658c10484bac23efe5c | | DeepSeek-VL-7B | 33.7 | Not recorded | 6a2f86e0d33dfb23e783be19 | | FGRPO | 28.42 | Not recorded | 6a210658c10484bac23efe5c | | FedAvg | 26.82 | Not recorded | 6a210658c10484bac23efe5c | | SCAFFOLD | 26.75 | Not recorded | 6a210658c10484bac23efe5c | | FedProx | 26.41 | Not recorded | 6a210658c10484bac23efe5c | ### VeriGeo-beam5 Total Accuracy: 82.74. Result ID: 6a2fa81efbc373592c2b8cc3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"8.7k","Inference":"beam5"}. This record does not assert independent reproduction. ### VeriGeo-greedy Total Accuracy: 79.31. Result ID: 6a2fa81efbc373592c2b8cc2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"8.7k","Inference":"greedy"}. This record does not assert independent reproduction. ### TR-CoT-Qwen2.5-VL-7B Total Accuracy: 79.2. Result ID: 6a2fa81efbc373592c2b8cc0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"65K + Geo170K"}. This record does not assert independent reproduction. ### GeoGen-SFT-7B Total Accuracy: 78. Result ID: 6a2fa81efbc373592c2b8cbf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"224K"}. This record does not assert independent reproduction. ### TR-CoT-InternVL2.5-8B Total Accuracy: 76.7. Result ID: 6a2fa81efbc373592c2b8cc1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8B","Data Scale":"65K + Geo170K"}. This record does not assert independent reproduction. ### TR-CoT-8B Total Accuracy: 75.9. Result ID: 6a2fa81efbc373592c2b8cbd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8B","Data Scale":"87K"}. This record does not assert independent reproduction. ### GeoGen-SFT-3B Total Accuracy: 75.6. Result ID: 6a2fa81efbc373592c2b8cbe. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"3B","Data Scale":"224K"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B-Instruct Total Accuracy: 72.07. Result ID: 6a2fa81efbc373592c2b8cb8. [Reporting paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"-"}. This record does not assert independent reproduction. ### MAVIS-7B Total Accuracy: 68.3. Result ID: 6a2fa81efbc373592c2b8cbc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"834K"}. This record does not assert independent reproduction. ### G-LLaVA-13B Total Accuracy: 67. Result ID: 6a2fa81efbc373592c2b8cbb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"13B","Data Scale":"170K"}. This record does not assert independent reproduction. ### G-LLaVA-7B Total Accuracy: 62.8. Result ID: 6a2fa81efbc373592c2b8cba. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"170K"}. This record does not assert independent reproduction. ### GPT-4o Total Accuracy: 61.4. Result ID: 6a2fa81efbc373592c2b8cb3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"Closed","Data Scale":"-"}. This record does not assert independent reproduction. ### InternVL2.5-8B Total Accuracy: 59. Result ID: 6a2fa81efbc373592c2b8cb7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8B","Data Scale":"-"}. This record does not assert independent reproduction. ### InternVL2-8B Total Accuracy: 56.5. Result ID: 6a2fa81efbc373592c2b8cb6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8B","Data Scale":"-"}. This record does not assert independent reproduction. ### Qwen2-VL-7B/8B Total Accuracy: 55.7. Result ID: 6a2fa81efbc373592c2b8cb5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B/8B","Data Scale":"-"}. This record does not assert independent reproduction. ### FGRPO Total Accuracy: 53.93. Result ID: 6a212fe1fbc373592c1c8948. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-4B"}. This record does not assert independent reproduction. ### FedProx Total Accuracy: 52.93. Result ID: 6a212fe1fbc373592c1c8946. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-4B"}. This record does not assert independent reproduction. ### SCAFFOLD Total Accuracy: 51.95. Result ID: 6a212fe1fbc373592c1c8947. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-4B"}. This record does not assert independent reproduction. ### FedAvg Total Accuracy: 51.4. Result ID: 6a212fe1fbc373592c1c8945. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-4B"}. This record does not assert independent reproduction. ### FGRPO Total Accuracy: 51.36. Result ID: 6a212fe1fbc373592c1c8944. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### FedProx Total Accuracy: 48.95. Result ID: 6a212fe1fbc373592c1c8942. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### FedAvg Total Accuracy: 48.87. Result ID: 6a212fe1fbc373592c1c8941. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### SCAFFOLD Total Accuracy: 48.42. Result ID: 6a212fe1fbc373592c1c8943. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Math-LLaVA-13B Total Accuracy: 47.8. Result ID: 6a2fa81efbc373592c2b8cb9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"13B","Data Scale":"360K"}. This record does not assert independent reproduction. ### GPT-4V Total Accuracy: 43.4. Result ID: 6a2fa81efbc373592c2b8cb2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"Closed","Data Scale":"-"}. This record does not assert independent reproduction. ### FGRPO Total Accuracy: 40.76. Result ID: 6a212fe1fbc373592c1c8940. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-3B"}. This record does not assert independent reproduction. ### FedProx Total Accuracy: 39.28. Result ID: 6a212fe1fbc373592c1c893e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-3B"}. This record does not assert independent reproduction. ### SCAFFOLD Total Accuracy: 38.08. Result ID: 6a212fe1fbc373592c1c893f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-3B"}. This record does not assert independent reproduction. ### FedAvg Total Accuracy: 35.59. Result ID: 6a212fe1fbc373592c1c893d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen2.5-3B"}. This record does not assert independent reproduction. ### DeepSeek-VL-7B Total Accuracy: 33.7. Result ID: 6a2fa81efbc373592c2b8cb4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"7B","Data Scale":"-"}. This record does not assert independent reproduction. ### FGRPO Total Accuracy: 28.42. Result ID: 6a212fe1fbc373592c1c894c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2-11B"}. This record does not assert independent reproduction. ### FedAvg Total Accuracy: 26.82. Result ID: 6a212fe1fbc373592c1c8949. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2-11B"}. This record does not assert independent reproduction. ### SCAFFOLD Total Accuracy: 26.75. Result ID: 6a212fe1fbc373592c1c894b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2-11B"}. This record does not assert independent reproduction. ### FedProx Total Accuracy: 26.41. Result ID: 6a212fe1fbc373592c1c894a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Llama-3.2-11B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.