Mathematical Reasoning on RBench-V Math
61.4AccuracyATLAS
Evaluation Results
| Method | Links | |
|---|---|---|
| ATLASEvaluation Protocol=Test-time scaling, Backbone=Gemini 3 Flash Preview2026.06 | 61.4 | |
| Pass@1Evaluation Protocol=Test-time scaling, Backbone=Gemini 3 Flash Preview2026.06 | 54.5 | |
| OpenAI o3Evaluation Protocol=Zero-shot2026.06 | 48.3 | |
| OpenAI o4-miniEvaluation Protocol=Zero-shot2026.06 | 43.2 | |
| Gemini 2.5 ProEvaluation Protocol=Zero-shot2026.06 | 42.6 | |
| Doubao-1.5-thinking-pro-mEvaluation Protocol=Zero-shot2026.06 | 38.6 | |
| OpenAI o1Evaluation Protocol=Zero-shot2026.06 | 34.7 | |
| GPT-OSS-120B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 30.1 | |
| GLM4.5-V-106B-A12B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 29.5 | |
| Qwen3-30B-A3B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 26.7 | |
| Skywork-OR1-32B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 25.3 | |
| GLM-4.5-Air-106B-A12B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 25 | |
| GPT-4oTikz-supplementation=false2026.03 | 24.4 | |
| Qwen3-VL-30B-A3B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 23.6 | |
| InternVL3.5-38B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 23.2 | |
| GLM4.5-V-106B-A12BTikz-supplementation=false2026.03 | 23 | |
| Qwen2.5-VL-32BTikz-supplementation=false2026.03 | 22.7 | |
| Qwen3-VL-30B-A3BTikz-supplementation=false2026.03 | 21.8 | |
| InternVL3.5-38BTikz-supplementation=false2026.03 | 21.6 | |
| DS-R1-Distill-Qwen-32B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 21.3 | |
| GLM-Z1-32B + GeoTikzBridge-BaseTikz-supplementation=true2026.03 | 20.1 | |
| InternVL3-78BTikz-supplementation=false2026.03 | 15.9 | |
| Qwen2.5-VL-72BTikz-supplementation=false2026.03 | 15.3 |