# Multimodal Deep Search on BC-VL Source: [Multimodal Deep Search on BC-VL]() Collection index: [llms.txt]() Data updated: 2026-05-12T16:52:01.780Z Evaluation of model performance on the BC-VL dataset for Multimodal Deep Search capabilities, 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 | Multimodal Deep Search | | Dataset (including any stated split) | BC-VL | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 37 of 37 | | 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 | | --- | --- | --- | --- | | GPT-5 | 57.6 | Not recorded | 6a02b4b0dbc836b87d824f68 | | MTA-DeepSearch-32B | 53.77 | Not recorded | 69d732707d0172140ec76a0c | | Gemini-3 Pro | 51.78 | Not recorded | 69d732707d0172140ec76a0c | | GPT-5 | 51.63 | Not recorded | 69d732707d0172140ec76a0c | | Claude-3.7-Sonnet | 50.4 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Gemini-2.5 Pro | 49.5 | Not recorded | 69d732707d0172140ec76a0c | | Claude-4-Sonnet | 48.6 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-30B + ODE-RL | 46.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | MTA-DeepSearch-8B | 44.36 | Not recorded | 69d732707d0172140ec76a0c | | Gemini-2.5 Pro | 43.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-30B + ODE-SFT | 43.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | MM-DeepResearch 32B | 43 | Not recorded | 69d732707d0172140ec76a0c | | GPT-5 | 42.9 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Gemini-2.5 Pro | 42.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-8B + ODE-RL | 41.9 | Not recorded | 6a02b4b0dbc836b87d824f68 | | GPT-5 | 41.6 | Not recorded | 69d732707d0172140ec76a0c | | Gemini-3-pro | 41.35 | Not recorded | 69d732707d0172140ec76a0c | | Gemini-2.5-pro | 39.85 | Not recorded | 69d732707d0172140ec76a0c | | Qwen3-VL-8B + ODE-SFT | 39.6 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-32B-Inst. | 38.69 | Not recorded | 69d732707d0172140ec76a0c | | Gemini-2.5 Flash | 38.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Gemini-2.5 Flash | 37.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-8B-Inst. | 35.89 | Not recorded | 69d732707d0172140ec76a0c | | Qwen3-VL-30B | 32.8 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Claude-3.7-Sonnet | 32.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-30B | 29.6 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Claude-4-Sonnet | 29.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-30B | 29.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Webwatcher-32B | 26.7 | Not recorded | 69d732707d0172140ec76a0c | | WebWatcher-32B | 26.7 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-8B | 26.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-8B | 25.1 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-32B-Inst. | 24.81 | Not recorded | 69d732707d0172140ec76a0c | | Qwen3-VL-8B | 24.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Webwatcher-7B | 20.3 | Not recorded | 69d732707d0172140ec76a0c | | WebWatcher-7B | 20.3 | Not recorded | 6a02b4b0dbc836b87d824f68 | | Qwen3-VL-8B-Inst. | 20.05 | Not recorded | 69d732707d0172140ec76a0c | ### GPT-5 Accuracy: 57.6. Result ID: 6a035a9647f34ce924e8f34a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### MTA-DeepSearch-32B Accuracy: 53.77. Result ID: 69d8334c47f34ce924bfdeeb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Ours","Model Scale":"32B"}. This record does not assert independent reproduction. ### Gemini-3 Pro Accuracy: 51.78. Result ID: 69d8334c47f34ce924bfdee4. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Agent Workflow"}. This record does not assert independent reproduction. ### GPT-5 Accuracy: 51.63. Result ID: 69d8334c47f34ce924bfdee2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Agent Workflow"}. This record does not assert independent reproduction. ### Claude-3.7-Sonnet Accuracy: 50.4. Result ID: 6a035a9647f34ce924e8f34c. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Gemini-2.5 Pro Accuracy: 49.5. Result ID: 69d8334c47f34ce924bfdee3. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Agent Workflow"}. This record does not assert independent reproduction. ### Claude-4-Sonnet Accuracy: 48.6. Result ID: 6a035a9647f34ce924e8f34b. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Qwen3-VL-30B + ODE-RL Accuracy: 46.1. Result ID: 6a035a9647f34ce924e8f358. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness","Training Strategy":"ODE-RL"}. This record does not assert independent reproduction. ### MTA-DeepSearch-8B Accuracy: 44.36. Result ID: 69d8334c47f34ce924bfdeea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Ours","Model Scale":"8B"}. This record does not assert independent reproduction. ### Gemini-2.5 Pro Accuracy: 43.1. Result ID: 6a035a9647f34ce924e8f346. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Qwen3-VL-30B + ODE-SFT Accuracy: 43.1. Result ID: 6a035a9647f34ce924e8f357. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness","Training Strategy":"ODE-SFT"}. This record does not assert independent reproduction. ### MM-DeepResearch 32B Accuracy: 43. Result ID: 69d8334c47f34ce924bfdee9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Multimodal Deep Search Agents","Model Scale":"32B"}. This record does not assert independent reproduction. ### GPT-5 Accuracy: 42.9. Result ID: 6a035a9647f34ce924e8f343. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Gemini-2.5 Pro Accuracy: 42.3. Result ID: 6a035a9647f34ce924e8f34d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Qwen3-VL-8B + ODE-RL Accuracy: 41.9. Result ID: 6a035a9647f34ce924e8f355. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness","Training Strategy":"ODE-RL"}. This record does not assert independent reproduction. ### GPT-5 Accuracy: 41.6. Result ID: 69d8334c47f34ce924bfdedd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Direct Answer"}. This record does not assert independent reproduction. ### Gemini-3-pro Accuracy: 41.35. Result ID: 69d8334c47f34ce924bfdedf. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Direct Answer"}. This record does not assert independent reproduction. ### Gemini-2.5-pro Accuracy: 39.85. Result ID: 69d8334c47f34ce924bfdede. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Direct Answer"}. This record does not assert independent reproduction. ### Qwen3-VL-8B + ODE-SFT Accuracy: 39.6. Result ID: 6a035a9647f34ce924e8f354. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness","Training Strategy":"ODE-SFT"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Inst. Accuracy: 38.69. Result ID: 69d8334c47f34ce924bfdee6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Agent Workflow","Model Scale":"32B"}. This record does not assert independent reproduction. ### Gemini-2.5 Flash Accuracy: 38.1. Result ID: 6a035a9647f34ce924e8f34e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Gemini-2.5 Flash Accuracy: 37.1. Result ID: 6a035a9647f34ce924e8f347. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Inst. Accuracy: 35.89. Result ID: 69d8334c47f34ce924bfdee5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Agent Workflow","Model Scale":"8B"}. This record does not assert independent reproduction. ### Qwen3-VL-30B Accuracy: 32.8. Result ID: 6a035a9647f34ce924e8f356. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness"}. This record does not assert independent reproduction. ### Claude-3.7-Sonnet Accuracy: 32.3. Result ID: 6a035a9647f34ce924e8f345. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Qwen3-VL-30B Accuracy: 29.6. Result ID: 6a035a9647f34ce924e8f349. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Claude-4-Sonnet Accuracy: 29.3. Result ID: 6a035a9647f34ce924e8f344. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Qwen3-VL-30B Accuracy: 29.3. Result ID: 6a035a9647f34ce924e8f350. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Webwatcher-32B Accuracy: 26.7. Result ID: 69d8334c47f34ce924bfdee8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Multimodal Deep Search Agents","Model Scale":"32B"}. This record does not assert independent reproduction. ### WebWatcher-32B Accuracy: 26.7. Result ID: 6a035a9647f34ce924e8f352. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Deep Search Agent"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Accuracy: 26.1. Result ID: 6a035a9647f34ce924e8f353. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Visual-Native Agent Harness"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Accuracy: 25.1. Result ID: 6a035a9647f34ce924e8f348. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Direct Reasoning"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Inst. Accuracy: 24.81. Result ID: 69d8334c47f34ce924bfdee1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Direct Answer","Model Scale":"32B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Accuracy: 24.3. Result ID: 6a035a9647f34ce924e8f34f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Evaluation Setting":"Agent Workflow"}. This record does not assert independent reproduction. ### Webwatcher-7B Accuracy: 20.3. Result ID: 69d8334c47f34ce924bfdee7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Multimodal Deep Search Agents","Model Scale":"7B"}. This record does not assert independent reproduction. ### WebWatcher-7B Accuracy: 20.3. Result ID: 6a035a9647f34ce924e8f351. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Setting":"Deep Search Agent"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Inst. Accuracy: 20.05. Result ID: 69d8334c47f34ce924bfdee0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Protocol":"Direct Answer","Model Scale":"8B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.