# Grounded Reasoning on MM-GCoT Source: [Grounded Reasoning on MM-GCoT]() Collection index: [llms.txt]() Data updated: 2026-05-29T11:43:44.080Z Evaluation of model performance on the MM-GCoT dataset, focusing on Grounded Reasoning capabilities measured by attribute, judgement, and object accuracy/consistency scores. 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 | Grounded Reasoning | | Dataset (including any stated split) | MM-GCoT | | Metric | Attribute Accuracy (A-Acc) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | 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 - [Attribute Accuracy (A-Acc)](): Higher is better; unit: PERCENT. - [Attribute Accuracy (G-Acc)](): Higher is better; unit: PERCENT. - [Attribute Consistency](): Higher is better; unit: PERCENT. - [Judgement Accuracy (A-Acc)](): Higher is better; unit: PERCENT. - [Judgement Accuracy (G-Acc)](): Higher is better; unit: PERCENT. - [Judgement Consistency](): Higher is better; unit: PERCENT. - [Object Accuracy (A-Acc)](): Higher is better; unit: PERCENT. - [Object Accuracy (G-Acc)](): Higher is better; unit: PERCENT. - [Object Consistency](): Higher is better; unit: PERCENT. - [Average Accuracy (A-Acc)](): Higher is better; unit: PERCENT. - [Average Accuracy (G-Acc)](): Higher is better; unit: PERCENT. - [Average Consistency](): Higher is better; unit: PERCENT. ## Recorded results | Method | Attribute Accuracy (A-Acc) | Result date | Report paper | | --- | --- | --- | --- | | GR3D-8B | 78.9 | Not recorded | 6a191dbfdbd7150506f82a2a | | ROVER-LSW-GRPO | 76.7 | Not recorded | 6a17cbf6dbd7150506f8119d | | ROVER-LSW | 75.6 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-13B GCoT | 74.7 | Not recorded | 6a17cbf6dbd7150506f8119d | | Qwen2.5-VL-7B | 73.6 | Not recorded | 6a191dbfdbd7150506f82a2a | | LLaVA-7B GCoT | 72.8 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-GCoT-7B | 72.8 | Not recorded | 6a191dbfdbd7150506f82a2a | | LLaVA-13B | 69.7 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-13B | 69.7 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-7B | 68.6 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-7B | 68.6 | Not recorded | 6a191dbfdbd7150506f82a2a | | InternVL2.5-8B | 61.9 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-7B | 59.7 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-7B | 59.7 | Not recorded | 6a191dbfdbd7150506f82a2a | | InternVL2.5-8B | 54.5 | Not recorded | 6a17cbf6dbd7150506f8119d | | Qwen2.5-VL-7B | 48.8 | Not recorded | 6a191dbfdbd7150506f82a2a | ### GR3D-8B Attribute Accuracy (A-Acc): 78.9. Result ID: 6a197be3fbc373592c14d947. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### ROVER-LSW-GRPO Attribute Accuracy (A-Acc): 76.7. Result ID: 6a17faccfbc373592c11a8eb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8.5B","Prompt Setting":"GCoT"}. This record does not assert independent reproduction. ### ROVER-LSW Attribute Accuracy (A-Acc): 75.6. Result ID: 6a17faccfbc373592c11a8ea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8.5B","Prompt Setting":"GCoT"}. This record does not assert independent reproduction. ### LLaVA-13B GCoT Attribute Accuracy (A-Acc): 74.7. Result ID: 6a17faccfbc373592c11a8e9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"13.1B","Prompt Setting":"GCoT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Attribute Accuracy (A-Acc): 73.6. Result ID: 6a197be3fbc373592c14d942. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Setting":"answer-first (AF)"}. This record does not assert independent reproduction. ### LLaVA-7B GCoT Attribute Accuracy (A-Acc): 72.8. Result ID: 6a17faccfbc373592c11a8e8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"6.8B","Prompt Setting":"GCoT"}. This record does not assert independent reproduction. ### LLaVA-GCoT-7B Attribute Accuracy (A-Acc): 72.8. Result ID: 6a197be3fbc373592c14d946. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LLaVA-13B Attribute Accuracy (A-Acc): 69.7. Result ID: 6a17faccfbc373592c11a8e6. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"13.1B","Prompt Setting":"AF"}. This record does not assert independent reproduction. ### LLaVA-13B Attribute Accuracy (A-Acc): 69.7. Result ID: 6a17faccfbc373592c11a8e7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"13.1B","Prompt Setting":"GF"}. This record does not assert independent reproduction. ### LLaVA-7B Attribute Accuracy (A-Acc): 68.6. Result ID: 6a17faccfbc373592c11a8e4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"6.8B","Prompt Setting":"AF"}. This record does not assert independent reproduction. ### LLaVA-7B Attribute Accuracy (A-Acc): 68.6. Result ID: 6a197be3fbc373592c14d944. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Setting":"answer-first (AF)"}. This record does not assert independent reproduction. ### InternVL2.5-8B Attribute Accuracy (A-Acc): 61.9. Result ID: 6a17faccfbc373592c11a8e2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.1B","Prompt Setting":"AF"}. This record does not assert independent reproduction. ### LLaVA-7B Attribute Accuracy (A-Acc): 59.7. Result ID: 6a17faccfbc373592c11a8e5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"6.8B","Prompt Setting":"GF"}. This record does not assert independent reproduction. ### LLaVA-7B Attribute Accuracy (A-Acc): 59.7. Result ID: 6a197be3fbc373592c14d945. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Setting":"grounding-first (GF)"}. This record does not assert independent reproduction. ### InternVL2.5-8B Attribute Accuracy (A-Acc): 54.5. Result ID: 6a17faccfbc373592c11a8e3. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.1B","Prompt Setting":"GF"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Attribute Accuracy (A-Acc): 48.8. Result ID: 6a197be3fbc373592c14d943. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Setting":"grounding-first (GF)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.