List your company

Commonsense Reasoning on Common sense QA (AUCOAA)

81.4AUCOAAAdaptive-Answer
46.66455.68264.773.718Jan 6, 2026

Evaluation results

MethodLinks
Adaptive-AnswerBackbone=Qwen3-8B2026.0181.4
Format-Adaptive-AnswerBackbone=Qwen3-8B2026.0181.3
Normalized-LengthBackbone=Qwen3-8B2026.0178.6
SFTBackbone=Qwen3-8B2026.0176.8
Hard-Length 8k → 4kBackbone=Qwen3-8B2026.0176.5
TWYNBackbone=Qwen3-8B2026.0174.8
Hard-Length 8kBackbone=Qwen3-8B2026.0173.3
Soft-LengthBackbone=Qwen3-8B2026.0173.3
Hard-Length 16kBackbone=Qwen3-8B2026.0172.5
Base modelBackbone=Qwen3-8B2026.0172.1
No-ThinkingBackbone=Qwen3-8B2026.0148