Share your thoughts, 1 month free Claude Pro on usSee more

Toxicity Evaluation on BOLD

0Toxic Rate

Base model

Updated 1mo ago

Evaluation Results

Method	Links
Base model 2026.01		0	0.031	0.024	3.35	-
RAD 2026.01		0	0.03	0.024	3.33	-
RAD 2026.01		0	0.026	0.021	3.39	-
RAD 2026.01		0	0.025	0.02	3.5	-
RAD 2026.01		0	0.024	0.02	3.64	-
RAD 2026.01		0	0.022	0.018	6.24	-
RAD 2026.01		0	0.023	0.02	11.58	-
SASA 2026.01		0	0.03	0.024	3.36	-
SASA 2026.01		0	0.031	0.025	3.37	-
SASA 2026.01		0	0.032	0.025	3.37	-
SASA 2026.01		0	0.032	0.025	3.37	-
SASA 2026.01		0	0.032	0.025	3.82	-
SASA 2026.01		0	0.033	0.026	4.93	-
TIDE 2026.01		0	0.025	0.021	3.35	1
Universal Self-consistency 2025.07		0	-	-	-	-
MAD 2025.07		0	-	-	-	-
MEP 2025.07		0	-	-	-	-
MetaCrit w/o ϕ↑ 2025.07		0	-	-	-	-
MetaCrit w/o ϕ↓ 2025.07		0	-	-	-	-
MetaCrit + GPT-3.5-Turbo 2025.07		0	-	-	-	-
MetaCrit + DeepSeek-v3 2025.07		0	-	-	-	-
MetaCrit + Claude-3.5-Sonnet 2025.07		0	-	-	-	-
MetaCrit + GPT-4o 2025.07		0	-	-	-	-
Self-refine 2025.07		6.4	-	-	-	-
ExpertPrompting 2025.07		12.9	-	-	-	-
Zero-shot-CoT 2025.07		16.3	-	-	-	-