List your company

Abstract Reasoning on ConceptARC

97.29AccuracyQwen3 + DIARC
-2.8123.177549.16575.1525Jun 25, 2026

Evaluation results

MethodLinks
Qwen3 + DIARC#Params=3.6B2026.0697.29
Qwen3 + SFT only#Params=3.6B2026.0696.46
Human (pass@1)#Params=-2026.0673
o4-mini#Params=N/A2026.0670.8
Gemini 2.5 Pro#Params=N/A2026.0666
Minitron + DIARC#Params=6.9B2026.0662.71
Minitron + SFT only#Params=6.9B2026.0660.42
Claude Sonnet 4#Params=N/A2026.0660.2
Llama-3.2 + DIARC#Params=2.6B2026.0659.17
Llama-3.2 + SFT only#Params=2.6B2026.0658.13
Qwen-2.5#Params=7B2026.0626
GPT-4#Params=N/A2026.0624
o4-mini (visual)#Params=N/A2026.068.1
Minitron#Params=8B2026.065.83
Qwen3#Params=4B2026.061.25
Llama-3.2#Params=3B2026.061.04