# Multi-task aerial image reasoning on HRVQA-VL (test) Source: [Multi-task aerial image reasoning on HRVQA-VL (test)]() Collection index: [llms.txt]() Data updated: 2026-05-07T08:09:06.118Z Evaluation of model performance on the test split of the HRVQA-VL dataset for multi-task aerial image reasoning, measured by Reasoning 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 | Multi-task aerial image reasoning | | Dataset (including any stated split) | HRVQA-VL (test) | | Metric | Reasoning Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 24 of 24 | | 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 - [Reasoning Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Reasoning Accuracy | Result date | Report paper | | --- | --- | --- | --- | | UAV-VL-R1 | 72.13 | Not recorded | 68fa7ab11e209131cea84a92 | | UAV-VL-R1 (Ours, all stages) | 72.13 | Not recorded | 68fa7ab11e209131cea84a92 | | UAV-VL-R1 | 68.94 | Not recorded | 68fa7ab11e209131cea84a92 | | UAV-VL-R1 (Ours, all stages) | 68.94 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct (SFT) | 52.6 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct (SFT) | 52.6 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-72B-Instruct | 51.46 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-72B-Instruct | 51.46 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-72B-Instruct | 46.67 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-72B-Instruct | 46.67 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-7B-Instruct | 43.83 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-7B-Instruct | 43.83 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-7B-Instruct | 38.4 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-7B-Instruct | 38.4 | Not recorded | 68fa7ab11e209131cea84a92 | | InstructBLIP-7B | 37.67 | Not recorded | 68fa7ab11e209131cea84a92 | | InstructBLIP-7B | 37.67 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct | 36.6 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct | 36.6 | Not recorded | 68fa7ab11e209131cea84a92 | | InstructBLIP-7B | 32.91 | Not recorded | 68fa7ab11e209131cea84a92 | | InstructBLIP-7B | 32.91 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct (SFT) | 26.93 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct (SFT) | 26.93 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct | 22.2 | Not recorded | 68fa7ab11e209131cea84a92 | | Qwen2-VL-2B-Instruct | 22.2 | Not recorded | 68fa7ab11e209131cea84a92 | ### UAV-VL-R1 Reasoning Accuracy: 72.13. Result ID: 69fc489747f34ce924dedd26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting","Stages":"all stages"}. This record does not assert independent reproduction. ### UAV-VL-R1 (Ours, all stages) Reasoning Accuracy: 72.13. Result ID: 69fc489747f34ce924dedd3b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### UAV-VL-R1 Reasoning Accuracy: 68.94. Result ID: 69fc489747f34ce924dedd25. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain","Stages":"all stages"}. This record does not assert independent reproduction. ### UAV-VL-R1 (Ours, all stages) Reasoning Accuracy: 68.94. Result ID: 69fc489747f34ce924dedd3a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct (SFT) Reasoning Accuracy: 52.6. Result ID: 69fc489747f34ce924dedd1d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct (SFT) Reasoning Accuracy: 52.6. Result ID: 69fc489747f34ce924dedd32. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-72B-Instruct Reasoning Accuracy: 51.46. Result ID: 69fc489747f34ce924dedd23. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-72B-Instruct Reasoning Accuracy: 51.46. Result ID: 69fc489747f34ce924dedd38. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-72B-Instruct Reasoning Accuracy: 46.67. Result ID: 69fc489747f34ce924dedd24. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-72B-Instruct Reasoning Accuracy: 46.67. Result ID: 69fc489747f34ce924dedd39. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-7B-Instruct Reasoning Accuracy: 43.83. Result ID: 69fc489747f34ce924dedd21. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-7B-Instruct Reasoning Accuracy: 43.83. Result ID: 69fc489747f34ce924dedd36. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-7B-Instruct Reasoning Accuracy: 38.4. Result ID: 69fc489747f34ce924dedd22. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-7B-Instruct Reasoning Accuracy: 38.4. Result ID: 69fc489747f34ce924dedd37. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### InstructBLIP-7B Reasoning Accuracy: 37.67. Result ID: 69fc489747f34ce924dedd1f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### InstructBLIP-7B Reasoning Accuracy: 37.67. Result ID: 69fc489747f34ce924dedd34. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct Reasoning Accuracy: 36.6. Result ID: 69fc489747f34ce924dedd1b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct Reasoning Accuracy: 36.6. Result ID: 69fc489747f34ce924dedd30. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Plain"}. This record does not assert independent reproduction. ### InstructBLIP-7B Reasoning Accuracy: 32.91. Result ID: 69fc489747f34ce924dedd20. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### InstructBLIP-7B Reasoning Accuracy: 32.91. Result ID: 69fc489747f34ce924dedd35. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct (SFT) Reasoning Accuracy: 26.93. Result ID: 69fc489747f34ce924dedd1e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct (SFT) Reasoning Accuracy: 26.93. Result ID: 69fc489747f34ce924dedd33. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct Reasoning Accuracy: 22.2. Result ID: 69fc489747f34ce924dedd1c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. ### Qwen2-VL-2B-Instruct Reasoning Accuracy: 22.2. Result ID: 69fc489747f34ce924dedd31. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Prompting Strategy":"Zero-shot Prompting"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.