List your company

Safety Evaluation on Tülu Safety Benchmarks 3

89DAN ScoreVCL
75.4878.9982.586.01Mar 4, 2026

Evaluation results

MethodLinks
VCLdecoding_settings=greedy decoding, temperature=none, max_length=5122026.038984.110086
Vector Ablationdecoding_settings=greedy decoding, temperature=none, max_length=5122026.0384809791
DROdecoding_settings=greedy decoding, temperature=none, max_length=5122026.0380729281
System Promptdecoding_settings=greedy decoding, temperature=none, max_length=5122026.0379759577
Llama-3.2-1B-SFTdecoding_settings=greedy decoding, temperature=none, max_length=5122026.0378749076
Self-CDdecoding_settings=greedy decoding, temperature=none, max_length=5122026.0376819183