List your company

Relational Reasoning on W-UP

98.3Accuracy (%)THINKLITE-VL
84.67688.21391.7595.287May 22, 2026

Evaluation results

MethodLinks
THINKLITE-VLMethod=THINKLITE-VL2026.0598.3
QWEN2.5-VL-7B + PGTBackbone=QWEN2.5-VL-7B, Data=PGT-augmented2026.0598
INTERNVL3-8B + PGTBackbone=INTERNVL3-8B, Data=PGT-augmented2026.0597.9
IMAGE JIGSAWMethod=IMAGE JIGSAW2026.0597.4
INTERNVL3-8BBackbone=INTERNVL3-8B, Data=Original2026.0597.2
QWEN2.5-VL-7BBackbone=QWEN2.5-VL-7B, Data=Original2026.0596.8
QWEN2.5-VL-7B + SPECIALIZED MIXBackbone=QWEN2.5-VL-7B, Data=Specialized Mix2026.0596.4
VIGORL-3BMethod=VIGORL-3B2026.0596.2
QWEN2.5-VL-3B + PGTBackbone=QWEN2.5-VL-3B, Data=PGT-augmented2026.0596.1
LLAVA-NEXT-LLAMA3-8BBackbone=LLAVA-NEXT-LLAMA3-8B, Data=Original2026.0593.8
LLAVA-NEXT-LLAMA3-8B + PGTBackbone=LLAVA-NEXT-LLAMA3-8B, Data=PGT-augmented2026.0593.8
QWEN2.5-VL-3BBackbone=QWEN2.5-VL-3B, Data=Original2026.0593.8
QWEN2.5-VL-3B + SPECIALIZED MIXBackbone=QWEN2.5-VL-3B, Data=Specialized Mix2026.0593.7
LLAVA-NEXT-7B + PGTBackbone=LLAVA-NEXT-7B, Data=PGT-augmented2026.0590.7
SPATIAL-LADDER-3BMethod=SPATIAL-LADDER-3B2026.0588.8
LLAVA-NEXT-7BBackbone=LLAVA-NEXT-7B, Data=Original2026.0585.2