List your company

Multi-risk safety monitoring on BeaverTails

80.1Accuracy (%)TELLME
14.16431.28248.465.518Feb 7, 2025

Evaluation results

MethodLinks
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0280.1
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0279.1
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0279
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0278.8
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0277.8
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0277.6
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0277.4
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0277.3
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0277
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0276.6
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0276.2
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0276.2
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0276.1
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0275.8
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0275.8
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0275.8
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0275
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0274.6
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0273
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0271.8
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0270.8
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0269.3
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0269.3
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0261.5
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0259.8
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0259.7
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0259.2
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0259.1
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0259
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0259
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0258.8
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0257.7
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0257.6
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0257.4
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0257.1
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0257
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0256.8
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0256.4
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0256.3
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0256.1
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0256.1
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0255.6
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0254.4
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0254.3
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0254.2
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0254.2
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0254.1
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0253.8
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0253.5
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0253.2
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0252.4
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0252.4
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0252.3
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0252.3
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0251.5
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0236.7
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0223
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0222.8
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0220.2
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0219.1
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0216.9
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0216.7