Gauri Kambhatla

CS PhD student @UTAustin. previously @umich 〽️. interested in #NLProc

[Gauri Kambhatla @gkambhat]

Thanks to my amazing collaborators: @Sanjana08395511, Angela Zhang, Alex Liu, Ravi Srinivasan, @jessyjli, and @mattlease!

[Gauri Kambhatla @gkambhat]

How can we consistently make LLM-generated distributions better align with the opinions of diverse population groups, and evaluate them robustly?

We study this in our paper, “Improving the Distributional Alignment of LLMs using Supervision”, to be presented at #ACL2026! 🧵

[Gauri Kambhatla @gkambhat]

For more results and analysis, please see our paper: https://t.co/n7Ozrnng79
Github: https://t.co/Wa1dd21iQC
Video: https://t.co/u0Ox3dWhsS
📍ACL Poster Session B: Sun July 5, 2-3:30p, San Diego

[Gauri Kambhatla @gkambhat]

In particular, we show how incorporating social science theories into a computational framework can help make LLM generations more aligned with human perceptions!
Data is available at https://t.co/qBy03aDOvR

[Gauri Kambhatla @gkambhat]

Excited to present our work (w/ @mattlease and @R_Ashwin) at #EMNLP2024 🌴on constructive deliberation and receptiveness! Stop by if you’re attending:
🗓️Poster Session C, Tues, Nov. 12th at 4pm
Paper: https://t.co/FqgB4tDJJV
Details in 🧵:

[Gauri Kambhatla @gkambhat]

We also analyze different aspects of receptiveness, as well as receptive reframing effectiveness for varying degrees of toxicity; we discuss how a tool based on our framework might be used for more flexible and creative content moderation.

[Gauri Kambhatla @gkambhat]

Come visit us for our FAccT presentation on 6/24 at 12:45pm UTC+9, or visit the virtual booth to learn more!
Paper: https://t.co/eHCyb2OVmD
Data: https://t.co/PNJRbhkOdm

[Gauri Kambhatla @gkambhat]

How do the differences in language use between people of different racial identities reflect known stereotypes? Excited to share our work (with @alethioguy and @radamihalcea) on Surfacing Racial Stereotypes through Identity Portrayal, in which we explore this question! #FAccT2022

[Gauri Kambhatla @gkambhat]

We surface underlying notions of racial saliency, a sense of in vs. out group and its relation to self-portrayal, and how visibility affects the way we think about other racial groups.