Risultati della ricerca - Datla, Vivek

Confidence-Based Response Abstinence: Improving LLM Trustworthiness via Activation-Based Uncertainty Estimation di Huang, Zhiqi, Datla, Vivek, Zhu, Chenyang, Samuel, Alfy, Liu, Daben, Kumar, Anoop, Soni, Ritesh

Testo

Alignment-Weighted DPO: A principled reasoning approach to improve safety alignment di Hu, Mengxuan, Datla, Vivek V., Kumar, Anoop, Guan, Zihan, Li, Sheng, Samuel, Alfy, Liu, Daben

Testo