List your company

Open-ended behavior generation on CAA

5.33CoAIS ScoreDiffMean
-0.131.28752.7054.1225Mar 6, 2026

Evaluation results

MethodLinks
DiffMeanBackbone=Llama-2-7b-hf2026.035.333.087.25.023.644.764.15
COLD-KernelBackbone=Llama-2-7b-hf2026.034.363.846.044.532.84.763.68
BaseBackbone=Llama-2-7b-hf2026.034.33.85.984.843.164.844.32
COLD-FDBackbone=Llama-2-7b-hf2026.033.942.587.225.184.54.364.06
ReFT(vector)Backbone=Llama-2-7b-hf2026.033.922.3675.383.884.664.24
COLD-FDBackbone=Llama-2-7b-chat-hf2026.030.825.063.322.624.926.21.23
BaseBackbone=Llama-2-7b-chat-hf2026.030.283.822.981.984.885.260.92
DiffMeanBackbone=Llama-2-7b-chat-hf2026.030.24.083.021.95.25.821.02
COLD-KernelBackbone=Llama-2-7b-chat-hf2026.030.23.863.32.225.25.40.96
ReFT(vector)Backbone=Llama-2-7b-chat-hf2026.030.083.722.662.244.985.540.69