List your company

Code Generation on LiveCodeBench (AUCOAA)

94.3AUCOAABase model
38.86853.25967.6582.041Jan 6, 2026

Evaluation results

MethodLinks
Base modelBackbone=Qwen3-8B2026.0194.3
Hard-Length 16kBackbone=Qwen3-8B2026.0194.1
SFTBackbone=Qwen3-8B2026.0193.9
TWYNBackbone=Qwen3-8B2026.0193.9
Format-Adaptive-AnswerBackbone=Qwen3-8B2026.0193.6
Hard-Length 8kBackbone=Qwen3-8B2026.0193.1
Adaptive-AnswerBackbone=Qwen3-8B2026.0193.1
Soft-LengthBackbone=Qwen3-8B2026.0193
Hard-Length 8k → 4kBackbone=Qwen3-8B2026.0192.7
Normalized-LengthBackbone=Qwen3-8B2026.0190
No-ThinkingBackbone=Qwen3-8B2026.0141