Error Detection on KITTI DK (noise-induced test set)
96RecallDETR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DETREvaluation Stage=Error Proposals2025.07 | 96 | 72 | 89 | 82 | |
| Llama (FT-CoT)Evaluation Stage=Error Validation, Mode=Fine-Tuned with Chain-of-Thought2025.07 | 96 | 94 | 92 | 95 | |
| AutoVDC (Llama (FT-CoT))Evaluation Stage=AutoVDC System, Backbone VLM=Llama (Fine-Tuned with CoT)2025.07 | 92 | 94 | 96 | 93 | |
| Llama (FT)Evaluation Stage=Error Validation, Mode=Fine-Tuned2025.07 | 91 | 91 | 87 | 91 | |
| AutoVDC (Llama (FT))Evaluation Stage=AutoVDC System, Backbone VLM=Llama (Fine-Tuned)2025.07 | 88 | 91 | 94 | 89 | |
| Gemini Flash 2.0Evaluation Stage=Error Validation2025.07 | 84 | 83 | 76 | 83 | |
| GPT 4.1Evaluation Stage=Error Validation2025.07 | 84 | 84 | 77 | 84 | |
| AutoVDC (Gemini Flash 2.0)Evaluation Stage=AutoVDC System, Backbone VLM=Gemini Flash 2.02025.07 | 80 | 83 | 90 | 81 | |
| AutoVDC (GPT 4.1)Evaluation Stage=AutoVDC System, Backbone VLM=GPT 4.12025.07 | 80 | 84 | 90 | 82 | |
| ViP-LLaVAEvaluation Stage=Error Validation2025.07 | 61 | 71 | 53 | 66 | |
| DeepSeek-VL2Evaluation Stage=Error Validation2025.07 | 58 | 99 | 69 | 73 | |
| AutoVDC (ViP-LLaVA)Evaluation Stage=AutoVDC System, Backbone VLM=ViP-LLaVA2025.07 | 58 | 71 | 82 | 64 | |
| AutoVDC (DeepSeek-VL2)Evaluation Stage=AutoVDC System, Backbone VLM=DeepSeek-VL22025.07 | 55 | 99 | 87 | 71 | |
| LlamaEvaluation Stage=Error Validation, Mode=Zero-shot2025.07 | 47 | 89 | 58 | 62 | |
| AutoVDC (Llama)Evaluation Stage=AutoVDC System, Backbone VLM=Llama (Zero-shot)2025.07 | 45 | 89 | 83 | 60 |