Visual Attribution on Thinking-Model Attribution Dataset General
0.73LDSvStream
Evaluation results
| Method | Links | ||
|---|---|---|---|
| vStreamModel=GLM-4.1V-9B-Thinking2026.04 | 0.73 | 0.95 | |
| AttnLRPModel=Cosmos-Reason1-7B2026.04 | 0.72 | 0.79 | |
| TAMModel=Qwen3-VL-8B-Thinking2026.04 | 0.71 | 0.89 | |
| InputGradModel=MiMo-VL-7B2026.04 | 0.71 | 0.68 | |
| vStreamModel=Qwen3-VL-8B-Thinking2026.04 | 0.7 | 1.02 | |
| AttnLRPModel=GLM-4.1V-9B-Thinking2026.04 | 0.7 | 0.75 | |
| vStreamModel=MiMo-VL-7B2026.04 | 0.7 | 1.1 | |
| TAMModel=Cosmos-Reason1-7B2026.04 | 0.69 | 1.01 | |
| vStreamModel=Cosmos-Reason1-7B2026.04 | 0.69 | 1.05 | |
| TAMModel=GLM-4.1V-9B-Thinking2026.04 | 0.68 | 1.08 | |
| AttnLRPModel=MiMo-VL-7B2026.04 | 0.66 | 1.04 | |
| AttnLRPModel=Qwen3-VL-8B-Thinking2026.04 | 0.65 | 0.98 | |
| InputGradModel=Qwen3-VL-8B-Thinking2026.04 | 0.64 | 0.75 | |
| TAMModel=MiMo-VL-7B2026.04 | 0.64 | 0.96 | |
| InputGradModel=GLM-4.1V-9B-Thinking2026.04 | 0.63 | 0.81 | |
| InputGradModel=Cosmos-Reason1-7B2026.04 | 0.6 | 0.75 | |
| AttentionModel=Qwen3-VL-8B-Thinking2026.04 | 0.43 | 0.26 | |
| AttentionModel=MiMo-VL-7B2026.04 | 0.39 | 0.28 | |
| AttentionModel=GLM-4.1V-9B-Thinking2026.04 | 0.38 | 0.27 | |
| AttentionModel=Cosmos-Reason1-7B2026.04 | 0.38 | 0.38 | |
| RandomModel=Qwen3-VL-8B-Thinking2026.04 | 0.29 | 0.07 | |
| RandomModel=GLM-4.1V-9B-Thinking2026.04 | 0.28 | 0.11 | |
| RandomModel=MiMo-VL-7B2026.04 | 0.28 | 0.12 | |
| RandomModel=Cosmos-Reason1-7B2026.04 | 0.27 | 0.12 |