List your company

Value Alignment on BeaverTails (test)

59.9Value Alignment ScoreNo Guidance
11.5424.09536.6549.205May 12, 2026

Evaluation results

MethodLinks
No GuidanceBackbone=QWEN2-1.5B2026.0559.9—
No GuidanceBackbone=LLAMA-3.2-3B2026.0558.95—
No GuidanceBackbone=MISTRAL-7B2026.0550.9—
System PromptBackbone=LLAMA-3.2-3B2026.0542.0428.7
System PromptBackbone=MISTRAL-7B2026.0541.418.7
ITIBackbone=LLAMA-3.2-3B2026.0540.6331.1
RE-ControlBackbone=LLAMA-3.2-3B2026.0539.2733.4
ITIBackbone=MISTRAL-7B2026.0537.226.9
No GuidanceBackbone=GPT-22026.0535.94—
DPO (LoRA)Backbone=LLAMA-3.2-3B2026.0534.7141.1
System PromptBackbone=QWEN2-1.5B2026.0534.4942.4
RE-ControlBackbone=QWEN2-1.5B2026.0534.2542.8
ITIBackbone=QWEN2-1.5B2026.0533.8443.5
System PromptBackbone=GPT-22026.0533.486.8
RE-ControlBackbone=MISTRAL-7B2026.0532.1236.9
DPO (LoRA)Backbone=QWEN2-1.5B2026.0530.5649
ITIBackbone=GPT-22026.0528.8519.7
SVGTBackbone=LLAMA-3.2-3B2026.0528.5851.5
SVGTBackbone=QWEN2-1.5B2026.0528.0453.2
DPO (LoRA)Backbone=MISTRAL-7B2026.0524.1652.5
RE-ControlBackbone=GPT-22026.0523.0735.8
DPO (LoRA)Backbone=GPT-22026.0520.6442.6
SVGTBackbone=GPT-22026.0519.8944.7
SVGTBackbone=MISTRAL-7B2026.0513.473.7