Mathematical Reasoning on Math Vista
65.3AccuracyBase
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BaseRatio=100%, Model Backbone=Qwen3-VL-4B2026.05 | 65.3 | 81.95 | 100 | |
| ERASERatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 62 | 78.98 | 96.38 | |
| Nemotron-Labs-Diffusion-VLM-8BGen. Mode=Linear SS, TPF=3.63 all samples 6.03 tok>100 7.45 tok>2002026.07 | 60.7 | — | — | |
| Nemotron-Labs-Diffusion-VLM-8BGen. Mode=AR, TPF=12026.07 | 60.4 | — | — | |
| LLaDA-V-8BGen. Mode=Diff., TPF=12026.07 | 59.7 | — | — | |
| Nemotron-Labs-Diffusion-VLM-8BGen. Mode=Diff., TPF=2.46 all samples 2.80 tok>100 3.15 tok>2002026.07 | 58.5 | — | — | |
| ERASERatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 57 | 77.1 | 94.09 | |
| DARTRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 53 | 69.49 | 84.8 | |
| IVC-PruneRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 52.8 | 76.81 | 93.74 | |
| ERASERatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 52.4 | 73.25 | 89.39 | |
| IVC-PruneRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 52.1 | 75.09 | 91.64 | |
| DARTRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 49.8 | 63.72 | 77.75 | |
| PruneSIDRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 49.8 | 63.72 | 77.75 | |
| IVC-PruneRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 48.9 | 66.15 | 80.73 | |
| PruneSIDRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 47.9 | 66.97 | 81.72 | |
| DARTRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 45.6 | 50.73 | 61.91 | |
| CDPrunerRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 45.1 | 60.84 | 74.24 | |
| LaViDaGen. Mode=Diff., TPF=12026.07 | 44.8 | — | — | |
| CDPrunerRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 44 | 55.15 | 67.3 | |
| PruneSIDRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 43.4 | 51.54 | 62.9 | |
| DimpleGen. Mode=Diff., TPF=12026.07 | 42.3 | — | — | |
| CDPrunerRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 39.2 | 43.71 | 53.34 | |
| MMaDAGen. Mode=Diff., TPF=12026.07 | 33.4 | — | — |