List your company

Code Generation on LiveCodeBench (Accuracy and Absolute Accuracy Change)

83.33AccuracySKILLGEN
19.338835.951952.56569.1781Feb 26, 2026Mar 16, 2026Apr 3, 2026Apr 21, 2026May 9, 2026May 27, 2026Jun 14, 2026

Evaluation results

MethodLinks
SKILLGENModel Name=Gemma-4-26B, Skill Augmentation Protocol=SKILL2026.0583.330
SKILLGENModel Name=GPT-5.4-Mini, Skill Augmentation Protocol=SKILL2026.0565.336
ReQATTQSQuantization Format=MXFP4, Precision=W4A162026.0654.52—
ReQATTQQuantization Format=MXFP4, Precision=W4A162026.0653.82—
BF16 FTQuantization Format=MXFP4, Precision=W4A162026.0653.68—
BF16 FTQuantization Format=NVFP4, Precision=W4A4KV42026.0653.68—
ReQATTQSQuantization Format=NVFP4, Precision=W4A4KV42026.0653.59—
ReQATTQQuantization Format=NVFP4, Precision=W4A4KV42026.0653.08—
QATQuantization Format=MXFP4, Precision=W4A162026.0652.01—
BF16 BaselineQuantization Format=MXFP4, Precision=W4A162026.0651.68—
BF16 BaselineQuantization Format=NVFP4, Precision=W4A4KV42026.0651.68—
QATQuantization Format=NVFP4, Precision=W4A4KV42026.0650.68—
PTQQuantization Format=MXFP4, Precision=W4A162026.0648.88—
PTQQuantization Format=NVFP4, Precision=W4A4KV42026.0647.99—
Dynamic-MAS + PRMBackbone=Qwen3-14B2026.0233—
Dynamic-MAS + Self-RefineBackbone=Qwen3-14B2026.0232.5—
Dynamic-MAS + ADv2Backbone=Qwen3-14B2026.0232—
Single AgentBackbone=Qwen3-14B2026.0229.5—
Dynamic-MAS + Multi-TAGBackbone=Qwen3-14B2026.0229.5—
Dynamic-MASBackbone=Qwen3-14B2026.0229—
OPDTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0625.1—
OPDTeacher-Student Setting=Qwen3-8B -> Llama3.1-8B-SFT2026.0624.8—
CDMTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0624.7—
SFTTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0624.3—
ALMTeacher-Student Setting=DeepSeek-R1 -> DeepSeek-R1-Distill-Qwen-7B2026.0624—
CDMTeacher-Student Setting=Qwen3-8B -> Llama3.1-8B-SFT2026.0622.9—
ALMTeacher-Student Setting=Qwen3-8B -> Llama3.1-8B-SFT2026.0622.4—
SFTTeacher-Student Setting=Qwen3-8B -> Llama3.1-8B-SFT2026.0621.8—