List your company

Theory of Mind for privacy control on ConfAIde Tier 3 (test)

95.444FR (E) Scorebase
71.0622477.3921283.72290.05188Feb 14, 2026

Evaluation results

MethodLinks
baseBackbone=Mistral-7B, Zero-shot transfer=true2026.0295.44488.4070.74188.40780.7041.48180.333
baseBackbone=Llama-3B, Zero-shot transfer=true2026.0287.59392.1480.3792.14894.3330.59394.333
PrivAct (Verifier)Backbone=Llama-3B, Zero-shot transfer=true2026.0286.03792.4070.3792.40791.9260.3791.926
PrivAct (Verifier)Backbone=Llama-8B, Zero-shot transfer=true2026.0285.48175.4443.51972.29651.4811.48150.37
PrivAct (Refiner)Backbone=Llama-3B, Zero-shot transfer=true2026.0285.3791.4810.3791.48192.1480.6392.148
PrivAct (Refiner)Backbone=Mistral-7B, Zero-shot transfer=true2026.0283.70476.5564.22272.48152.371.6350.741
baseBackbone=Llama-8B, Zero-shot transfer=true2026.0283.3776.4074.44472.33352.0741.6350.444
PrivAct (Refiner)Backbone=Llama-8B, Zero-shot transfer=true2026.0282.3774.0371.48172.92654.7410.66754.444
PrivAct (Verifier)Backbone=Mistral-7B, Zero-shot transfer=true2026.028077.3334.44473.25955.1851.44453.963
PrivAct (Verifier)Backbone=Qwen-4B, Zero-shot transfer=true2026.0274.40771.2961.3769.96363.1850.51962.963
PrivAct (Refiner)Backbone=Qwen-4B, Zero-shot transfer=true2026.0272.25976.4071.6375.07464.9260.88964.407
baseBackbone=Qwen-4B, Zero-shot transfer=true2026.027273.4442.74170.70461.3330.55661.185