List your company

LLM Steering on BASE→USER.NEG (out-of-distribution)

2.195SteerabilityL2S
0.888761.227881.5671.90612Apr 4, 2026

Evaluation results

MethodLinks
L2SModel=Llama-2-7B-Chat, Layer Selection Strategy=W2S2026.042.19593.2
L2SModel=Llama-2-7B-Chat, Layer Selection Strategy=Fixed Layer2026.041.97789.9
L2SModel=Qwen1.5-14B-Chat, Layer Selection Strategy=W2S2026.041.31674.8
CAAModel=Llama-2-7B-Chat, Layer Selection Strategy=W2S2026.041.17274.1
CAAModel=Qwen1.5-14B-Chat, Layer Selection Strategy=W2S2026.041.15470.9
L2SModel=Qwen1.5-14B-Chat, Layer Selection Strategy=Fixed Layer2026.041.0672.2
CAAModel=Llama-2-7B-Chat, Layer Selection Strategy=Fixed Layer2026.041.02870.1
CAAModel=Qwen1.5-14B-Chat, Layer Selection Strategy=Fixed Layer2026.040.93968.6