# Causal Spatial Reasoning on CausalSpatial Level 2 Source: [Causal Spatial Reasoning on CausalSpatial Level 2]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:39:31.160Z Evaluation of model performance on the CausalSpatial Level 2 dataset, focusing on causal and spatial reasoning capabilities measured by various anticipation scores. 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 | Causal Spatial Reasoning | | Dataset (including any stated split) | CausalSpatial Level 2 | | Metric | Collision Anticipation | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [Collision Anticipation](): Higher is better; unit: PERCENT. - [Compatibility Anticipation](): Higher is better; unit: PERCENT. - [Occlusion Anticipation](): Higher is better; unit: PERCENT. - [Trajectory Anticipation](): Higher is better; unit: PERCENT. - [Average Anticipation Score](): Higher is better; unit: PERCENT. - [Not Sure Rate (NSR)](): Lower is better; unit: PERCENT. ## Recorded results | Method | Collision Anticipation | Result date | Report paper | | --- | --- | --- | --- | | Human | 78.57 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Gemini2.5 Flash | 42.31 | Not recorded | 69705ddae25ad2fa9d1ed67a | | GPT5 | 40.94 | Not recorded | 69705ddae25ad2fa9d1ed67a | | GPT5-mini | 35.43 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-30B-A3B-T. | 35.09 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Gemini2.5 Pro | 30.65 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen2.5-VL-7B-I. | 30.4 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-30B-A3B-I. | 30.16 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Sonnet 3.7 | 29.41 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Spatial-MLLM | 24.41 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-4B-I. | 24.41 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-8B-T. | 23.75 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-8B-I. | 22.76 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Qwen3-VL-2B-I. | 19.35 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Spatial-VLM | 15.75 | Not recorded | 69705ddae25ad2fa9d1ed67a | | LLaVAOneVision1.5-8B-I. | 15.38 | Not recorded | 69705ddae25ad2fa9d1ed67a | | Random | 8.54 | Not recorded | 69705ddae25ad2fa9d1ed67a | ### Human Collision Anticipation: 78.57. Result ID: 6997a7bf5698665ec4fddbed. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini2.5 Flash Collision Anticipation: 42.31. Result ID: 6997a7bf5698665ec4fddbeb. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT5 Collision Anticipation: 40.94. Result ID: 6997a7bf5698665ec4fddbe8. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT5-mini Collision Anticipation: 35.43. Result ID: 6997a7bf5698665ec4fddbe9. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-VL-30B-A3B-T. Collision Anticipation: 35.09. Result ID: 6997a7bf5698665ec4fddbe7. [Reporting paper]() Recorded evaluation configuration: {"Scale":"30B-A3B","Reasoning\_Trained":true}. This record does not assert independent reproduction. ### Gemini2.5 Pro Collision Anticipation: 30.65. Result ID: 6997a7bf5698665ec4fddbea. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen2.5-VL-7B-I. Collision Anticipation: 30.4. Result ID: 6997a7bf5698665ec4fddbe1. [Reporting paper]() Recorded evaluation configuration: {"Scale":"7B"}. This record does not assert independent reproduction. ### Qwen3-VL-30B-A3B-I. Collision Anticipation: 30.16. Result ID: 6997a7bf5698665ec4fddbe6. [Reporting paper]() Recorded evaluation configuration: {"Scale":"30B-A3B"}. This record does not assert independent reproduction. ### Sonnet 3.7 Collision Anticipation: 29.41. Result ID: 6997a7bf5698665ec4fddbec. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Spatial-MLLM Collision Anticipation: 24.41. Result ID: 6997a7bf5698665ec4fddbdf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-VL-4B-I. Collision Anticipation: 24.41. Result ID: 6997a7bf5698665ec4fddbe3. [Reporting paper]() Recorded evaluation configuration: {"Scale":"4B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-T. Collision Anticipation: 23.75. Result ID: 6997a7bf5698665ec4fddbe5. [Reporting paper]() Recorded evaluation configuration: {"Scale":"8B","Reasoning\_Trained":true}. This record does not assert independent reproduction. ### Qwen3-VL-8B-I. Collision Anticipation: 22.76. Result ID: 6997a7bf5698665ec4fddbe4. [Reporting paper]() Recorded evaluation configuration: {"Scale":"8B"}. This record does not assert independent reproduction. ### Qwen3-VL-2B-I. Collision Anticipation: 19.35. Result ID: 6997a7bf5698665ec4fddbe2. [Reporting paper]() Recorded evaluation configuration: {"Scale":"2B"}. This record does not assert independent reproduction. ### Spatial-VLM Collision Anticipation: 15.75. Result ID: 6997a7bf5698665ec4fddbde. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LLaVAOneVision1.5-8B-I. Collision Anticipation: 15.38. Result ID: 6997a7bf5698665ec4fddbe0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Scale":"8B"}. This record does not assert independent reproduction. ### Random Collision Anticipation: 8.54. Result ID: 6997a7bf5698665ec4fddbee. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.