# Visual Perception on SEED-Bench Image Source: [Visual Perception on SEED-Bench Image]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:29:49.240Z Evaluation of general visual perception capabilities using the SEED-Bench Image dataset, measured by 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 Perception | | Dataset (including any stated split) | SEED-Bench Image | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 18 of 18 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | MG-LLaVA | 73.7 | Not recorded | 68fa73d61e209131cea5e688 | | GPT-4V | 72.3 | Not recorded | 68fa73d61e209131cea5e688 | | SPHINX-1k | 71.6 | Not recorded | 68fa73d61e209131cea5e688 | | MG-LLaVA | 71.5 | Not recorded | 68fa73d61e209131cea5e688 | | MG-LLaVA | 70.8 | Not recorded | 68fa73d61e209131cea5e688 | | GeminiPro Vision | 70.7 | Not recorded | 68fa73d61e209131cea5e688 | | MG-LLaVA | 70.2 | Not recorded | 68fa73d61e209131cea5e688 | | MG-LLaVA | 69.4 | Not recorded | 68fa73d61e209131cea5e688 | | SPHINX | 69.1 | Not recorded | 68fa73d61e209131cea5e688 | | LLaVA-1.5 | 68.2 | Not recorded | 68fa73d61e209131cea5e688 | | MiniCPM-V2 | 67.1 | Not recorded | 68fa73d61e209131cea5e688 | | LLaVA-1.5 | 66.1 | Not recorded | 68fa73d61e209131cea5e688 | | Qwen-VL-Plus | 65.7 | Not recorded | 68fa73d61e209131cea5e688 | | LLaVA-HR | 64.5 | Not recorded | 68fa73d61e209131cea5e688 | | Qwen-VL-Chat | 58.2 | Not recorded | 68fa73d61e209131cea5e688 | | Qwen-VL | 56.3 | Not recorded | 68fa73d61e209131cea5e688 | | InstructBLIP | 53.4 | Not recorded | 68fa73d61e209131cea5e688 | | BLIP-2 | 46.4 | Not recorded | 68fa73d61e209131cea5e688 | ### MG-LLaVA Accuracy: 73.7. Result ID: 69930e2d3a6c31484913626d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Yi1.5-34B","Params.":"34.4B","Res.":"336 (768)"}. This record does not assert independent reproduction. ### GPT-4V Accuracy: 72.3. Result ID: 69930e2d3a6c31484913625c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### SPHINX-1k Accuracy: 71.6. Result ID: 69930e2d3a6c314849136266. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","Params.":"10B","Res.":"448"}. This record does not assert independent reproduction. ### MG-LLaVA Accuracy: 71.5. Result ID: 69930e2d3a6c31484913626b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"LLaMA3-8B","Params.":"8.4B","Res.":"336 (768)"}. This record does not assert independent reproduction. ### MG-LLaVA Accuracy: 70.8. Result ID: 69930e2d3a6c31484913626c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-13B","Params.":"13.6B","Res.":"336 (768)"}. This record does not assert independent reproduction. ### GeminiPro Vision Accuracy: 70.7. Result ID: 69930e2d3a6c31484913625d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### MG-LLaVA Accuracy: 70.2. Result ID: 69930e2d3a6c314849136269. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Phi3-3.8B","Params.":"4.2B","Res.":"336 (768)"}. This record does not assert independent reproduction. ### MG-LLaVA Accuracy: 69.4. Result ID: 69930e2d3a6c31484913626a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","Params.":"7.4B","Res.":"336 (768)"}. This record does not assert independent reproduction. ### SPHINX Accuracy: 69.1. Result ID: 69930e2d3a6c314849136265. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","Params.":"10B","Res.":"224"}. This record does not assert independent reproduction. ### LLaVA-1.5 Accuracy: 68.2. Result ID: 69930e2d3a6c314849136264. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-13B","Params.":"13.4B","Res.":"336"}. This record does not assert independent reproduction. ### MiniCPM-V2 Accuracy: 67.1. Result ID: 69930e2d3a6c314849136267. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"LLM":"MiniCPM-2.4B","Params.":"2.8B","Res.":"448"}. This record does not assert independent reproduction. ### LLaVA-1.5 Accuracy: 66.1. Result ID: 69930e2d3a6c314849136263. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","Params.":"7.2B","Res.":"336"}. This record does not assert independent reproduction. ### Qwen-VL-Plus Accuracy: 65.7. Result ID: 69930e2d3a6c31484913625e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LLaVA-HR Accuracy: 64.5. Result ID: 69930e2d3a6c314849136268. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-13B","Params.":"13.4B","Res.":"448 (1024)"}. This record does not assert independent reproduction. ### Qwen-VL-Chat Accuracy: 58.2. Result ID: 69930e2d3a6c314849136262. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen-7B","Params.":"9.6B","Res.":"448"}. This record does not assert independent reproduction. ### Qwen-VL Accuracy: 56.3. Result ID: 69930e2d3a6c314849136261. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen-7B","Params.":"9.6B","Res.":"448"}. This record does not assert independent reproduction. ### InstructBLIP Accuracy: 53.4. Result ID: 69930e2d3a6c314849136260. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","Params.":"8.2B","Res.":"224"}. This record does not assert independent reproduction. ### BLIP-2 Accuracy: 46.4. Result ID: 69930e2d3a6c31484913625f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-13B","Params.":"14.2B","Res.":"224"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.