# Multimodal Question Answering on MME-RealWorld-Lite 1.0 (test) Source: [Multimodal Question Answering on MME-RealWorld-Lite 1.0 (test)]() Collection index: [llms.txt]() Data updated: 2026-03-10T10:07:23.065Z Evaluation of Multimodal Question Answering performance on the test split of the MME-RealWorld-Lite 1.0 dataset, reporting detailed perception and reasoning accuracies across various sub-tasks. 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 | Multimodal Question Answering | | Dataset (including any stated split) | MME-RealWorld-Lite 1.0 (test) | | Metric | Perception (AD) Acc | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 19 of 19 | | 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 - [Perception (AD) Acc](): Higher is better; unit: PERCENT. - [Perception (MO) Acc](): Higher is better; unit: PERCENT. - [Perception (OCR) Acc](): Higher is better; unit: PERCENT. - [Perception (RS) Acc](): Higher is better; unit: PERCENT. - [Perception (DT) Acc](): Higher is better; unit: PERCENT. - [Reasoning (AD) Acc](): Higher is better; unit: PERCENT. - [Reasoning (DT) Acc](): Higher is better; unit: PERCENT. - [Reasoning (OCR) Acc](): Higher is better; unit: PERCENT. - [Reasoning (MO) Acc](): Higher is better; unit: PERCENT. - [Overall Acc](): Higher is better; unit: PERCENT. ## Recorded results | Method | Perception (AD) Acc | Result date | Report paper | | --- | --- | --- | --- | | HART-7B | 57.7 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B (MGPO + SFT) | 55.4 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B (GRPO + SFT) | 49.7 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B (MGPO) | 44 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B (SFT) | 43.7 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B (GRPO) | 43.7 | Not recorded | 69a519822d603e6de3e3c99b | | Claude 3.5 Sonnet | 40.8 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-32B | 40.7 | Not recorded | 69a519822d603e6de3e3c99b | | InternVL3-38B | 40 | Not recorded | 69a519822d603e6de3e3c99b | | LLaVA-OneVision-72B | 40 | Not recorded | 69a519822d603e6de3e3c99b | | LLaVA-OneVision-7B | 39.4 | Not recorded | 69a519822d603e6de3e3c99b | | InternVL3-8B | 36.9 | Not recorded | 69a519822d603e6de3e3c99b | | DeepEyes-7B | 33.4 | Not recorded | 69a519822d603e6de3e3c99b | | Pixel-Reasoner-7B | 30.9 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-72B | 30.6 | Not recorded | 69a519822d603e6de3e3c99b | | Qwen2.5-VL-7B | 30 | Not recorded | 69a519822d603e6de3e3c99b | | Gemini-1.5-Pro | 26.6 | Not recorded | 69a519822d603e6de3e3c99b | | GPT-4o-mini | 24.2 | Not recorded | 69a519822d603e6de3e3c99b | | GPT-4o | 22.4 | Not recorded | 69a519822d603e6de3e3c99b | ### HART-7B Perception (AD) Acc: 57.7. Result ID: 69afed506dbb7075b958d747. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Model Category":"Visual Grounded Reasoning Models"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (MGPO + SFT) Perception (AD) Acc: 55.4. Result ID: 69afed506dbb7075b958d744. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Post-Training Strategy":"MGPO + SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (GRPO + SFT) Perception (AD) Acc: 49.7. Result ID: 69afed506dbb7075b958d742. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"7B","Post-Training Strategy":"GRPO + SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (MGPO) Perception (AD) Acc: 44. Result ID: 69afed506dbb7075b958d743. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Post-Training Strategy":"MGPO"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (SFT) Perception (AD) Acc: 43.7. Result ID: 69afed506dbb7075b958d740. [Reporting paper]() Recorded evaluation configuration: {"Parameters":"7B","Post-Training Strategy":"SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B (GRPO) Perception (AD) Acc: 43.7. Result ID: 69afed506dbb7075b958d741. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"7B","Post-Training Strategy":"GRPO"}. This record does not assert independent reproduction. ### Claude 3.5 Sonnet Perception (AD) Acc: 40.8. Result ID: 69afed506dbb7075b958d74f. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Private Models"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B Perception (AD) Acc: 40.7. Result ID: 69afed506dbb7075b958d748. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"32B","Model Category":"Larger-scale Open-source Models"}. This record does not assert independent reproduction. ### InternVL3-38B Perception (AD) Acc: 40. Result ID: 69afed506dbb7075b958d749. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"38B","Model Category":"Larger-scale Open-source Models"}. This record does not assert independent reproduction. ### LLaVA-OneVision-72B Perception (AD) Acc: 40. Result ID: 69afed506dbb7075b958d74b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"72B","Model Category":"Larger-scale Open-source Models"}. This record does not assert independent reproduction. ### LLaVA-OneVision-7B Perception (AD) Acc: 39.4. Result ID: 69afed506dbb7075b958d73e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"7B","Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### InternVL3-8B Perception (AD) Acc: 36.9. Result ID: 69afed506dbb7075b958d73f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"8B","Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### DeepEyes-7B Perception (AD) Acc: 33.4. Result ID: 69afed506dbb7075b958d746. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Model Category":"Visual Grounded Reasoning Models"}. This record does not assert independent reproduction. ### Pixel-Reasoner-7B Perception (AD) Acc: 30.9. Result ID: 69afed506dbb7075b958d745. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Model Category":"Visual Grounded Reasoning Models"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Perception (AD) Acc: 30.6. Result ID: 69afed506dbb7075b958d74a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"72B","Model Category":"Larger-scale Open-source Models"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Perception (AD) Acc: 30. Result ID: 69afed506dbb7075b958d73d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"7B","Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### Gemini-1.5-Pro Perception (AD) Acc: 26.6. Result ID: 69afed506dbb7075b958d74d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Private Models"}. This record does not assert independent reproduction. ### GPT-4o-mini Perception (AD) Acc: 24.2. Result ID: 69afed506dbb7075b958d74c. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Private Models"}. This record does not assert independent reproduction. ### GPT-4o Perception (AD) Acc: 22.4. Result ID: 69afed506dbb7075b958d74e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Private Models"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.