List your company

Misalignment detection on Taylor

99AccuracyTELLME
52.7264.73576.7588.765Feb 7, 2025

Evaluation results

MethodLinks
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0299
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0299
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0296.3
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0296.3
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0294.3
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0294
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0293.5
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0292.8
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0292.8
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0291.5
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0291.3
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0291.3
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0291
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0290.8
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0290.8
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0290.8
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0289
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0288.8
TELLMEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0288.5
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0287.8
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0286.2
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0286
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0286
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0285.8
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0285.5
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Judge2025.0285.2
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0284.5
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Linear Probe2025.0283.5
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0283.3
TELLMEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0282
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0280.5
OriginBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0280.5
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0280.3
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0280
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0279.8
SAE-Multi-TopKBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0279.8
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0278.8
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Self-Sim2025.0278.5
Skip TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0278.2
SAE-KLBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0278
SAEBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0277.8
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Judge2025.0277.8
Skip TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0276.8
OriginBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0276.5
SAEBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0276.3
SAE-KLBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0276.3
TranscoderBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0276.3
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Linear Probe2025.0275.8
SAE-Multi-TopKBackbone=Llama-3.1-8B, Evaluation Protocol=Self-Sim2025.0275.5
TranscoderBackbone=Qwen2.5-7B, Evaluation Protocol=Linear Probe2025.0275.5
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0269.3
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0269.3
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0268.8
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0268.8
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0268.8
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Self-Sim2025.0267.5
TELLMEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0263
SAEBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0257.8
SAE-KLBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0257.5
SAE-Multi-TopKBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0257
OriginBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0256.3
Skip TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0254.8
TranscoderBackbone=Mistral-7B-v0.3, Evaluation Protocol=Judge2025.0254.5