# Road Scene on RoadSceneBench Source: [Road Scene on RoadSceneBench]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:58:29.656Z Evaluation of model performance on general road scene understanding tasks using the RoadSceneBench dataset, reporting Precision and Recall. 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 | Road Scene | | Dataset (including any stated split) | RoadSceneBench | | Metric | Precision | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 36 of 36 | | 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 - [Precision](): Higher is better; unit: PERCENT. - [Recall](): Higher is better; unit: PERCENT. ## Recorded results | Method | Precision | Result date | Report paper | | --- | --- | --- | --- | | MapVLM | 82.8 | Not recorded | 692d217bd6b0dd3342931cba | | MapVLM | 81.62 | Not recorded | 692d217bd6b0dd3342931cba | | MapVLM | 75.78 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-8B | 73.07 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-8B | 72.79 | Not recorded | 692d217bd6b0dd3342931cba | | MapVLM | 72.14 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-3B | 71.78 | Not recorded | 692d217bd6b0dd3342931cba | | ERNIE-4.5-VL-28B-A3B | 69.98 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-7B | 69.79 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-78B | 69.02 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-32B | 68.07 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-72B | 67.85 | Not recorded | 692d217bd6b0dd3342931cba | | GPT-4o | 67.64 | Not recorded | 692d217bd6b0dd3342931cba | | Gemini-2.5-pro | 65.91 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-32B | 65.53 | Not recorded | 692d217bd6b0dd3342931cba | | Claude-3.7V Sonnet | 61.58 | Not recorded | 692d217bd6b0dd3342931cba | | Gemini-2.5-pro | 60.61 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-8B | 57.34 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-78B | 55.53 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-32B | 53.54 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-1B | 52.67 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-2B | 52.59 | Not recorded | 692d217bd6b0dd3342931cba | | GPT-4o | 51.79 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-72B | 50.57 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-32B | 50.4 | Not recorded | 692d217bd6b0dd3342931cba | | ERNIE-4.5-VL-28B-A3B | 50.11 | Not recorded | 692d217bd6b0dd3342931cba | | Claude-3.7V Sonnet | 47.3 | Not recorded | 692d217bd6b0dd3342931cba | | LLaVA-Onevision / Next-7B | 42.8 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-7B | 42.67 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen2.5-VL-3B | 40.61 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-8B | 39.24 | Not recorded | 692d217bd6b0dd3342931cba | | LLaVA-Onevision / Next-7B | 38.65 | Not recorded | 692d217bd6b0dd3342931cba | | Qwen3-VL-2B | 36.15 | Not recorded | 692d217bd6b0dd3342931cba | | DeepSeek-VL2 | 28.5 | Not recorded | 692d217bd6b0dd3342931cba | | InternVL3-1B | 26.11 | Not recorded | 692d217bd6b0dd3342931cba | | DeepSeek-VL2 | 16.63 | Not recorded | 692d217bd6b0dd3342931cba | ### MapVLM Precision: 82.8. Result ID: 6995e2ba3badfaffe6ea5ebd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Strategy":"SFT+HRRP-T"}. This record does not assert independent reproduction. ### MapVLM Precision: 81.62. Result ID: 6995e2ba3badfaffe6ea5ebc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Strategy":"SFT"}. This record does not assert independent reproduction. ### MapVLM Precision: 75.78. Result ID: 6995e2ba3badfaffe6ea5ee1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Strategy":"SFT+HRRP-T"}. This record does not assert independent reproduction. ### InternVL3-8B Precision: 73.07. Result ID: 6995e2ba3badfaffe6ea5eb3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Precision: 72.79. Result ID: 6995e2ba3badfaffe6ea5eba. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### MapVLM Precision: 72.14. Result ID: 6995e2ba3badfaffe6ea5ee0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Strategy":"SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Precision: 71.78. Result ID: 6995e2ba3badfaffe6ea5eb5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### ERNIE-4.5-VL-28B-A3B Precision: 69.98. Result ID: 6995e2ba3badfaffe6ea5eaf. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Precision: 69.79. Result ID: 6995e2ba3badfaffe6ea5eb6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### InternVL3-78B Precision: 69.02. Result ID: 6995e2ba3badfaffe6ea5eb4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B Precision: 68.07. Result ID: 6995e2ba3badfaffe6ea5eb7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Precision: 67.85. Result ID: 6995e2ba3badfaffe6ea5eb8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### GPT-4o Precision: 67.64. Result ID: 6995e2ba3badfaffe6ea5eac. [Reporting paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### Gemini-2.5-pro Precision: 65.91. Result ID: 6995e2ba3badfaffe6ea5ead. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-32B Precision: 65.53. Result ID: 6995e2ba3badfaffe6ea5ebb. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Claude-3.7V Sonnet Precision: 61.58. Result ID: 6995e2ba3badfaffe6ea5eae. [Reporting paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### Gemini-2.5-pro Precision: 60.61. Result ID: 6995e2ba3badfaffe6ea5ed1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Precision: 57.34. Result ID: 6995e2ba3badfaffe6ea5ede. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### InternVL3-78B Precision: 55.53. Result ID: 6995e2ba3badfaffe6ea5ed8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-32B Precision: 53.54. Result ID: 6995e2ba3badfaffe6ea5edf. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### InternVL3-1B Precision: 52.67. Result ID: 6995e2ba3badfaffe6ea5eb2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-2B Precision: 52.59. Result ID: 6995e2ba3badfaffe6ea5eb9. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### GPT-4o Precision: 51.79. Result ID: 6995e2ba3badfaffe6ea5ed0. [Reporting paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-72B Precision: 50.57. Result ID: 6995e2ba3badfaffe6ea5edc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B Precision: 50.4. Result ID: 6995e2ba3badfaffe6ea5edb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### ERNIE-4.5-VL-28B-A3B Precision: 50.11. Result ID: 6995e2ba3badfaffe6ea5ed3. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Claude-3.7V Sonnet Precision: 47.3. Result ID: 6995e2ba3badfaffe6ea5ed2. [Reporting paper]() Recorded evaluation configuration: {"Access":"Closed-Source"}. This record does not assert independent reproduction. ### LLaVA-Onevision / Next-7B Precision: 42.8. Result ID: 6995e2ba3badfaffe6ea5eb1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Precision: 42.67. Result ID: 6995e2ba3badfaffe6ea5eda. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Precision: 40.61. Result ID: 6995e2ba3badfaffe6ea5ed9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### InternVL3-8B Precision: 39.24. Result ID: 6995e2ba3badfaffe6ea5ed7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### LLaVA-Onevision / Next-7B Precision: 38.65. Result ID: 6995e2ba3badfaffe6ea5ed5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### Qwen3-VL-2B Precision: 36.15. Result ID: 6995e2ba3badfaffe6ea5edd. [Reporting paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### DeepSeek-VL2 Precision: 28.5. Result ID: 6995e2ba3badfaffe6ea5eb0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### InternVL3-1B Precision: 26.11. Result ID: 6995e2ba3badfaffe6ea5ed6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. ### DeepSeek-VL2 Precision: 16.63. Result ID: 6995e2ba3badfaffe6ea5ed4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Access":"Open-Source"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.