Shruti Joshi, a PhD student working under the supervision of Dhanya Sridhar, Assistant Professor in the Department of Computer Science and Operations Research (DIRO) at Université de Montréal, Core Academic Member at Mila, and Canada CIFAR AI Chair, has been awarded the 2025 Antidote Scholarship in Natural Language Processing (NLP), granted by Druide informatique.
Shruti Joshi’s work focuses on developing interpretation, control and reliability methods for large language models (LLMs). She created the Sparse Shift Autoencoder method, which helps control LLM activations in order to reduce bias in the generated texts, among other things. Before joining Professor Sridhar’s team, Joshi was a research programmer at the Max Planck Institute in Tübingen, Germany, where she contributed to various publications on modular neural architectures.
Valued at $20,000, the Antidote Scholarship in NLP is awarded annually to a student in Université de Montréal’s Department of Computer Science and Operations Research (DIRO) whose work is affiliated with Mila. The scholarship is funded by the $1 million Druide Fund for Research in Text Analysis, which was established to celebrate the twentieth anniversary of the Antidote software and the fiftieth anniversary of DIRO.
