Relational Reasoning on W-UP
98.3Accuracy (%)THINKLITE-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| THINKLITE-VLMethod=THINKLITE-VL2026.05 | 98.3 | |
| QWEN2.5-VL-7B + PGTBackbone=QWEN2.5-VL-7B, Data=PGT-augmented2026.05 | 98 | |
| INTERNVL3-8B + PGTBackbone=INTERNVL3-8B, Data=PGT-augmented2026.05 | 97.9 | |
| IMAGE JIGSAWMethod=IMAGE JIGSAW2026.05 | 97.4 | |
| INTERNVL3-8BBackbone=INTERNVL3-8B, Data=Original2026.05 | 97.2 | |
| QWEN2.5-VL-7BBackbone=QWEN2.5-VL-7B, Data=Original2026.05 | 96.8 | |
| QWEN2.5-VL-7B + SPECIALIZED MIXBackbone=QWEN2.5-VL-7B, Data=Specialized Mix2026.05 | 96.4 | |
| VIGORL-3BMethod=VIGORL-3B2026.05 | 96.2 | |
| QWEN2.5-VL-3B + PGTBackbone=QWEN2.5-VL-3B, Data=PGT-augmented2026.05 | 96.1 | |
| LLAVA-NEXT-LLAMA3-8BBackbone=LLAVA-NEXT-LLAMA3-8B, Data=Original2026.05 | 93.8 | |
| LLAVA-NEXT-LLAMA3-8B + PGTBackbone=LLAVA-NEXT-LLAMA3-8B, Data=PGT-augmented2026.05 | 93.8 | |
| QWEN2.5-VL-3BBackbone=QWEN2.5-VL-3B, Data=Original2026.05 | 93.8 | |
| QWEN2.5-VL-3B + SPECIALIZED MIXBackbone=QWEN2.5-VL-3B, Data=Specialized Mix2026.05 | 93.7 | |
| LLAVA-NEXT-7B + PGTBackbone=LLAVA-NEXT-7B, Data=PGT-augmented2026.05 | 90.7 | |
| SPATIAL-LADDER-3BMethod=SPATIAL-LADDER-3B2026.05 | 88.8 | |
| LLAVA-NEXT-7BBackbone=LLAVA-NEXT-7B, Data=Original2026.05 | 85.2 |