Kanishka Misra

Kanishka Misra

Research Assistant Professor at Toyota Technological Institute at Chicago

Speech & Language @ TTIC

UT Austin CompLing Group

I am a Research Assistant Professor at the Toyota Technological Institute at Chicago, a philanthropically endowed academic computer science institute located on the University of Chicago campus.

My research program lies at the intersection of Cognitive Science, Linguistics, and Artificial Intelligence. I am primarily interested in characterizing the statistical mechanisms that underlie the acquisition and generalization of complex linguistic phenomena and conceptual meaning. To this end, I: (1) develop methods to evaluate and analyze AI models from the perspective of semantic cognition; and (2) use AI models as simulated learners to test and generate novel hypotheses about language acquisition and generalization. My research has been recognized with awards at EACL 2023, ACL 2023, and EMNLP 2024!

Previously, I was a postdoctoral fellow in the Linguistics department at UT Austin, working with Dr. Kyle Mahowald. Before that, I was a PhD student at Purdue University, where I worked on Natural Language Understanding with Dr. Julia Taylor Rayz at the AKRaNLU Lab. I also worked closely with Dr. Allyson Ettinger and her lab at UChicago.

My email is kanishka [at] ttic [dot] edu.[why is it like that?]

Note: I am not currently looking to hire any PhD students, but please reach out to me if I can be of assistance or point you to other fantastic researchers who are looking for students. I will unfortunately not respond to your email if its content suggests you have not read my website.

Other things:

  • I co-organize the UChicago/TTIC NLP Seminar, along with the wonderful Mina Lee and Zhewei Sun. We are constantly looking for external speakers – please reach out if you are interested to present! I strongly encourage people from underrepresented and marginalized groups to reach out.

  • I am the author of minicons, a python package that facilitates large scale behavioral analyses of transformer language models.

  • I spent Fall 2022 as a Research Intern at Google AI working on multi-hop reasoning and language models.

Recent News

SEE ALL

Recent Publications

SEE ALL

Quickly discover relevant content by filtering publications.

Characterizing the Role of Similarity in the Property Inferences of Language Models

Behavioral and Causal analysis of the role of similarity and taxonomic relations in Language Models’ Property Inferences.
Juan Diego Rodriguez, Aaron Mueller, Kanishka Misra
Code arxiv
Characterizing the Role of Similarity in the Property Inferences of Language Models

Generating novel experimental hypotheses from language models: A case study on cross-dative generalization

Investigating LMs as simulated learners to understand the conditions that enable cross-dative generalization and derive novel hypotheses to test on human learners.
Kanishka Misra, Najoung Kim
Code arxiv
Generating novel experimental hypotheses from language models: A case study on cross-dative generalization

Language Models Learn Rare Phenomena from Less Rare Phenomena: The Case of the Missing AANNs

Analysis of how statistical learners of language acquire rare linguistic phenomena such as the article+adjective+numeral+noun construction.
Kanishka Misra, Kyle Mahowald
Code Poster arxiv ACL Anthology
Language Models Learn Rare Phenomena from Less Rare Phenomena: The Case of the Missing AANNs

Experimental Contexts Can Facilitate Robust Semantic Property Inference in Language Models, but Inconsistently

Rigorous analyses of the extent to which LMs can improve their ability to make property inferences for novel concepts in the presence of experimental contexts.)
Kanishka Misra, Allyson Ettinger, Kyle Mahowald
Code Poster arxiv ACL Anthology
Experimental Contexts Can Facilitate Robust Semantic Property Inference in Language Models, but Inconsistently

Abstraction via exemplars? A representational case study on lexical category inference in BERT

Work on understanding the representational dynamics underlying category-abstraction in LMs such as BERT.
Kanishka Misra, Najoung Kim
PDF Poster
Abstraction via exemplars? A representational case study on lexical category inference in BERT

Recent Posts

Introducing minicons: Running large scale behavioral analyses on transformer language models

In this post, I showcase my new python library that implements simple computations to facilitate large-scale evaluation of transformer language models.
Last updated on Mar 24, 2021 10 min read research
Introducing $\texttt{minicons}$: Running large scale behavioral analyses on transformer language models

Contact

Note: I am not currently looking to hire any PhD students, but please reach out to me if I can be of assistance or point you to other fantastic researchers who are looking for students. I will unfortunately not respond to your email if its content suggests you have not read my website.