List your company

Truncated CoT Answering on BBEH

0.665AOCFaithfulness Only
0.496520.540260.5840.62774Feb 18, 2026

Evaluation results

MethodLinks
Faithfulness OnlyBackbone=Qwen3-14B2026.020.665
MAT-SteerBackbone=Qwen3-14B2026.020.644
REMULBackbone=Qwen3-14B2026.020.626
OriginalBackbone=Qwen3-14B2026.020.58
Balanced RewardsBackbone=Qwen3-14B2026.020.574
Correctness OnlyBackbone=Qwen3-14B2026.020.518
Hint OptimizedBackbone=Qwen3-14B2026.020.503