一作ICLR 2026Poster· Social Aspects / Privacy-preserving Statistics and Machine Learning
DP-Fusion: Token-Level Differentially Private Inference for Large Language Models
Rushil Thareja · Preslav Nakov · Praneeth Vepakomma · Nils Lukas
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Rushil Thareja · Preslav Nakov · Praneeth Vepakomma · Nils Lukas
Zhuohan Xie · Daniil Orel · Rushil Thareja · Dhruv Sahnan · Hachem Madmoun · Fan Zhang · Debopriyo Banerjee · Georgi Nenkov Georgiev · Xueqing Peng · Lingfei Qian · Jimin Huang · Jinyan Su · Aaryamonvikram Singh · Rui Xing · Rania Elbadry · Chen Xu · Haonan Li · Fajri Koto · Ivan Koychev · Tanmoy Chakraborty · Yuxia Wang · Salem Lahlou · Veselin Stoyanov · Sophia Ananiadou · Preslav Nakov
Multi-step symbolic reasoning is essential for robust financial analysis; yet, current benchmarks largely overlook this capability. Existing datasets such as FinQA and ConvFinQA emphasize final numerical answers while neglecting the intermediate reasoning steps required for transparency and verif...