List your company

Safety Alignment on Safety BeaverTails, HEx-PHI (test)

95.67BeaverTails ScoreDASE
82.555685.960389.36592.7697Jan 29, 2026

Evaluation results

MethodLinks
DASEBase Model=GPT-5.2, Size=Undisclosed2026.0195.6710097.84
DASEBase Model=Kimi-K2-Instruct, Size=1T2026.0194.7910097.4
GPT-5.2Size=Undisclosed2026.0194.3310097.16
Gemma-2 + SaRFTSize=9B2026.0194.1695.294.68
Kimi-K2-InstructSize=1T2026.0193.9498.3396.13
Qwen3-235B-A22B-Instruct-2507Size=235B2026.0193.839996.41
Qwen3-32BSize=32B2026.0191.6299.6795.64
Qwen2.5 + SaRFTSize=7B2026.0189.9887.888.89
gpt-oss-120bSize=120B2026.0188.9998.3393.66
LLaMA-3 + SaRFTSize=8B2026.0183.0685.6784.36