Arnavi Chheda-Kothary retweetledi
Arnavi Chheda-Kothary
129 posts
Arnavi Chheda-Kothary
@arnavic
PhD student @ University of Washington CSE | Previously @ Ai2, MSFT, Columbia, UW
Katılım Mart 2010
174 Takip Edilen209 Takipçiler
Arnavi Chheda-Kothary retweetledi
Congratulations to the Super Bowl champion @Seahawks! This defense was special. MVP Kenneth Walker was dominant. And Sam Darnold gave us one of the best comeback stories in a long time. Enjoy the celebration.
English
English
Belated update: I defended my PhD last month!
I am tremendously grateful to my advisors, @HannaHajishirzi and @YejinChoinka. Without their incredible support, I wouldn’t have had so much fun exploring bold ideas, like taking a journey into the ocean of LLM pretraining data. 🥰🥰
English
It’s been incredible for me to return to Ai2 as a student researcher on this project with @turingmusician @lucyluwang and @josephcc, getting to combine my background in accessibility and engineering to bring this prototype to life! Please feel free to reach out with feedback.
English
Paper+Figure QA is early and still evolving, so expect some rough edges. We’d love your feedback as we improve it! Paper+Figure QA is only available for open access papers, so there may be some papers we do not currently support. In those cases, please try a different paper.
English
Ever want to ask questions about a paper, including its figures & tables? 📊📈 Want smoother interactions w/papers on desktop & mobile?
Try Paper+Figure QA, a new tool from @allen_ai that answers with the original figures, tables, and excerpts from papers: paperfigureqa.allen.ai
English
Arnavi Chheda-Kothary retweetledi
Agent benchmarks don't measure true *AI* advances
We built one that's hard & trustworthy
👉AstaBench tests agents w/ *standardized tools* on 2400+ scientific research problems
👉SOTA results across 22 agent *classes*
👉AgentBaselines agents suite
🆕arxiv.org/abs/2510.21652
🧵👇
English
Arnavi Chheda-Kothary retweetledi
@umsi we developed accessibility guidelines to evaluate AI coding tools like @Replit , @julesagent and @code . Blind or visually impaired developers, #A11Y experts and accademics, please participate in our study to ensure programming remains accessible!
idea11y.dev/VibeCheck/
English
English
Thank you to my collaborators, and a huge huge thank you to @BrianHCI who has stuck with me and with this work through several rejection cycles. This work happened as a part of my Master’s thesis working at the CEAL lab, and I’m grateful that we saw it over the finish line! (6/6)
English
Read more about our work here — dl.acm.org/doi/10.1145/37…. For those at CHI, I’ll be presenting on Wednesday during the Spatial Interactions session! Please come find me then or anytime this week to chat further about all things screen readers, accessibility, and AI! (5/6)
English
While other mediums like mobile devices have spatial affordances built in, desktops continue to feel linear and homogenous when parsing content with screen readers. (3/6)
English
In this work, we built and deployed a browser extension to enable spatial input and spatial output based screen reader interactions for increased layout understanding of web pages on desktops. (2/6)
English
Kicking off my very first CHI today! Great to be at #CHI2025 in beautiful Yokohama, where I’ll be presenting work on incorporating spatial interactions into desktop screen readers titled: "It Brought Me Joy": Opportunities for Spatial Browsing in Desktop Screen Readers. (1/6)
English
It was such a great experience to have been a part of bringing this to the Ai2 Playground! Congrats to the whole team 🎉
Ai2@allen_ai
For years it’s been an open question — how much is a language model learning and synthesizing information, and how much is it just memorizing and reciting? Introducing OLMoTrace, a new feature in the Ai2 Playground that begins to shed some light. 🔦
English
Today we're unveiling OLMoTrace, a tool that enables everyone to understand the outputs of LLMs by connecting to their training data.
We do this on unprecedented scale and in real time: finding matching text between model outputs and 4 trillion training tokens within seconds. ✨
Ai2@allen_ai
For years it’s been an open question — how much is a language model learning and synthesizing information, and how much is it just memorizing and reciting? Introducing OLMoTrace, a new feature in the Ai2 Playground that begins to shed some light. 🔦
English
A huge thanks to all my collaborators, and special shoutout to my wonderful advisors @wobbrockjo and @jonfroehlich.
English
In this work, we present a Human-AI system designed to facilitate deeper engagement and conversations between blind or low vision (BLV) family members and their sighted children around child-created visual artwork.
English
Excited to be at #IUI25 in Cagliari, Italy this week to present our paper “ArtInsight: Enabling AI-Powered Artwork Engagement for Mixed Visual-Ability Families”. You can find the full paper here: dl.acm.org/doi/full/10.11…
English