List your company

Safe-or-harmful binary classification on BeaverTails

84.6AccuracyTELLME
51.52860.11468.777.286Feb 7, 2025

Evaluation results

MethodLinks
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0284.6
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0284.6
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0284.5
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0284.4
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0284.4
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0284.2
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0283.8
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0283.1
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0282.9
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0282.7
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0282.5
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0282.5
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0282.5
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0282.5
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0282.3
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0282.3
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0282.1
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0282.1
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0282
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0281.6
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0281.5
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0281.3
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0281.2
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0276.9
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0275.6
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0275.5
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0275.4
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0275.3
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0275.2
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0275.2
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0275.2
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0275.1
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0274.8
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0274.7
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0274.7
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0274.6
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273.8
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273.7
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273.4
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273.3
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0273.1
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0273.1
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0273
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0272.9
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0272.8
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0272.7
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0272.1
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0272
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0271.9
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0271.8
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0271.8
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0271.8
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0271.6
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0271.6
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0270.4
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0270.4
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0265
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0263.7
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0255.6
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0255.2
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0253.6
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0253.4
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0252.8