List your company

Alignment Evaluation on Sycophancy

51Mean ImprovementiPASa
29.1634.8340.546.17Sep 25, 2025

Evaluation results

MethodLinks
iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0951500
PASfModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0951500
iPASaModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0943410
PASfModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0943420
iPASaModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0933290
iPASwoModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0933290
PASfModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0933290
iPASwoModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0933280
iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0930210