List your company

Grade-school Reasoning on GSM8K Aug

59.7AccuracyCoT-SFT
7.28420.89234.548.108Jun 29, 2026

Evaluation results

MethodLinks
CoT-SFTBackbone=Qwen3-VL-2B-Instruct2026.0659.7131.4
CoT-SFTBackbone=LLaMA-3.2-1B2026.0654.1123.2
DLR (Full)Backbone=Qwen3-VL-2B-Instruct2026.0653.46
DLR (Full)Backbone=LLaMA-3.2-1B2026.0643.46.4
ReGuLaRBackbone=LLaMA-3.2-1B2026.0634.93.7
CoLaRBackbone=Qwen3-VL-2B-Instruct2026.0628.35.7
CoLaRBackbone=LLaMA-3.2-1B2026.0626.65.6
RoTBackbone=Qwen3-VL-2B-Instruct2026.0623.332
CoconutBackbone=LLaMA-3.2-1B2026.0620.56
iCoTBackbone=LLaMA-3.2-1B2026.0619.80
Direct-SFTBackbone=Qwen3-VL-2B-Instruct2026.0615.60
Direct-SFTBackbone=LLaMA-3.2-1B2026.0613.90
CoconutBackbone=Qwen3-VL-2B-Instruct2026.0613.46
CODIBackbone=LLaMA-3.2-1B2026.0613.36
CODIBackbone=Qwen3-VL-2B-Instruct2026.069.36