# Multi-entity reasoning grounding on ME-RSRG (test) Source: [Multi-entity reasoning grounding on ME-RSRG (test)]() Collection index: [llms.txt]() Data updated: 2026-03-16T07:18:54.440Z Evaluation of multi-entity reasoning grounding performance on the test split of the ME-RSRG dataset, reporting subject accuracy, object accuracy, and mean accuracy at 0.5 IoU threshold. 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-entity reasoning grounding | | Dataset (including any stated split) | ME-RSRG (test) | | Metric | Subject Accuracy (IoU=0.5) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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 - [Subject Accuracy (IoU=0.5)](): Higher is better; unit: PERCENT. - [Object Accuracy (IoU=0.5)](): Higher is better; unit: PERCENT. - [Mean Accuracy (IoU=0.5)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Subject Accuracy (IoU=0.5) | Result date | Report paper | | --- | --- | --- | --- | | Qwen2.5-VL-7B | 44.4 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-8B | 39.08 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2.5-VL-3B | 38.63 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-4B | 38.52 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-4B | 38.02 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-8B | 37.96 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-4B | 36.39 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-8B | 35.33 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-4B | 34.49 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-8B | 33.76 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2.5-VL-7B | 31.75 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2-VL-2B | 27.55 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2.5-VL-3B | 27.04 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-8B | 26.82 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2-VL-2B | 23.18 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen3-VL-4B | 20.95 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-OneVision | 18.14 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2.5-VL-7B | 11.31 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-OneVision | 9.8 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL2.5-4B | 9.18 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-1.5-7b | 7.95 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2.5-VL-3B | 7.45 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL2.5-4B | 6.94 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-1.5-7b | 3.42 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-8B | 2.35 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL2.5-4B | 1.68 | Not recorded | 69b78e690636e5bf4afbdac8 | | Qwen2-VL-2B | 1.4 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-1.5-7b | 1.29 | Not recorded | 69b78e690636e5bf4afbdac8 | | InternVL3.5-4B | 0.5 | Not recorded | 69b78e690636e5bf4afbdac8 | | LLaVA-OneVision | 0 | Not recorded | 69b78e690636e5bf4afbdac8 | ### Qwen2.5-VL-7B Subject Accuracy (IoU=0.5): 44.4. Result ID: 69b7ad2247f34ce9249f3b19. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### InternVL3.5-8B Subject Accuracy (IoU=0.5): 39.08. Result ID: 69b7ad2247f34ce9249f3b28. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Subject Accuracy (IoU=0.5): 38.63. Result ID: 69b7ad2247f34ce9249f3b16. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen3-VL-4B Subject Accuracy (IoU=0.5): 38.52. Result ID: 69b7ad2247f34ce9249f3b1c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### InternVL3.5-4B Subject Accuracy (IoU=0.5): 38.02. Result ID: 69b7ad2247f34ce9249f3b25. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Subject Accuracy (IoU=0.5): 37.96. Result ID: 69b7ad2247f34ce9249f3b1f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen3-VL-4B Subject Accuracy (IoU=0.5): 36.39. Result ID: 69b7ad2247f34ce9249f3b1b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### InternVL3.5-8B Subject Accuracy (IoU=0.5): 35.33. Result ID: 69b7ad2247f34ce9249f3b27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### InternVL3.5-4B Subject Accuracy (IoU=0.5): 34.49. Result ID: 69b7ad2247f34ce9249f3b24. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Subject Accuracy (IoU=0.5): 33.76. Result ID: 69b7ad2247f34ce9249f3b1e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Subject Accuracy (IoU=0.5): 31.75. Result ID: 69b7ad2247f34ce9249f3b18. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### Qwen2-VL-2B Subject Accuracy (IoU=0.5): 27.55. Result ID: 69b7ad2247f34ce9249f3b13. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Subject Accuracy (IoU=0.5): 27.04. Result ID: 69b7ad2247f34ce9249f3b15. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Subject Accuracy (IoU=0.5): 26.82. Result ID: 69b7ad2247f34ce9249f3b1d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### Qwen2-VL-2B Subject Accuracy (IoU=0.5): 23.18. Result ID: 69b7ad2247f34ce9249f3b12. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### Qwen3-VL-4B Subject Accuracy (IoU=0.5): 20.95. Result ID: 69b7ad2247f34ce9249f3b1a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### LLaVA-OneVision Subject Accuracy (IoU=0.5): 18.14. Result ID: 69b7ad2247f34ce9249f3b2e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Subject Accuracy (IoU=0.5): 11.31. Result ID: 69b7ad2247f34ce9249f3b17. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### LLaVA-OneVision Subject Accuracy (IoU=0.5): 9.8. Result ID: 69b7ad2247f34ce9249f3b2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### InternVL2.5-4B Subject Accuracy (IoU=0.5): 9.18. Result ID: 69b7ad2247f34ce9249f3b22. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### LLaVA-1.5-7b Subject Accuracy (IoU=0.5): 7.95. Result ID: 69b7ad2247f34ce9249f3b2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"EAR"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Subject Accuracy (IoU=0.5): 7.45. Result ID: 69b7ad2247f34ce9249f3b14. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### InternVL2.5-4B Subject Accuracy (IoU=0.5): 6.94. Result ID: 69b7ad2247f34ce9249f3b21. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### LLaVA-1.5-7b Subject Accuracy (IoU=0.5): 3.42. Result ID: 69b7ad2247f34ce9249f3b2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"SFT"}. This record does not assert independent reproduction. ### InternVL3.5-8B Subject Accuracy (IoU=0.5): 2.35. Result ID: 69b7ad2247f34ce9249f3b26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### InternVL2.5-4B Subject Accuracy (IoU=0.5): 1.68. Result ID: 69b7ad2247f34ce9249f3b20. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### Qwen2-VL-2B Subject Accuracy (IoU=0.5): 1.4. Result ID: 69b7ad2247f34ce9249f3b11. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### LLaVA-1.5-7b Subject Accuracy (IoU=0.5): 1.29. Result ID: 69b7ad2247f34ce9249f3b29. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### InternVL3.5-4B Subject Accuracy (IoU=0.5): 0.5. Result ID: 69b7ad2247f34ce9249f3b23. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. ### LLaVA-OneVision Subject Accuracy (IoU=0.5): 0. Result ID: 69b7ad2247f34ce9249f3b2c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Optimization":"Baseline"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.