Kento Nishi｜AI Researcher, LiveTL+HyperChat Dev🐔
@kento_nishi
20-year-old AI researcher and web developer at Harvard University.
🚨 ICML 2025 Paper! 🚨 Excited to announce "Representation Shattering in Transformers: A Synthetic Study with Knowledge Editing." 🔗 arxiv.org/abs/2410.17194 We uncover a new phenomenon, Representation Shattering, to explain why KE edits negatively affect LLMs' reasoning. 🧵👇
Thank you to everyone who swung by our poster presentation!!! So many engaging conversations today. #ICML2025
🚨 ICML 2025 Paper! 🚨 Excited to announce "Representation Shattering in Transformers: A Synthetic Study with Knowledge Editing." 🔗 arxiv.org/abs/2410.17194 We uncover a new phenomenon, Representation Shattering, to explain why KE edits negatively affect LLMs' reasoning. 🧵👇
Q. Why does editing knowledge sometimes reduce overall performance? A. Representation shatters! Great insights by a super start undergrad @kento_nishi, supervised expertly by @EkdeepL, with @MayaOkawa, @RahulRam3sh, and Mikail Khona.
New paper---freshly accepted to ICML! Detailed thread coming soon, but pretty excited about this project. We use synthetic knowledge graphs to study why knowledge editing protocols can screw up model capabilities, finding what we call a "representation shattering" effect!
Paper link: arxiv.org/abs/2410.17194 Shoutout especially to our monster undergrad @kento_nishi who led this project, and all collaborators! @MayaOkawa, @RahulRam3sh, Mikail Khona, and @Hidenori8Tanaka
New paper---freshly accepted to ICML! Detailed thread coming soon, but pretty excited about this project. We use synthetic knowledge graphs to study why knowledge editing protocols can screw up model capabilities, finding what we call a "representation shattering" effect!
Accepted to ICML 2025!🎉See you in Vancouver! arxiv.org/abs/2410.17194
2000-day streak! github.com/KentoNishi
“No government—regardless of which party is in power—should dictate what private universities can teach, whom they can admit and hire, and which areas of study and inquiry they can pursue.” - President Alan Garber hrvd.me/GarberRespond3…
TLDR: Given an in-context task, and as context is scaled, LMs can form ‘in-context representations’ that reflect the task. Team: @corefpark @EkdeepL @YongyiYang7 @MayaOkawa @kento_nishi @wattenberg @Hidenori8Tanaka Special thanks: @ndif_team arxiv.org/pdf/2501.00070 2/N
This project was a true collaborative effort where everyone contributed to major parts of the project! Big thanks to the team! @a_jy_l, @EkdeepL, @YongyiYang7, @MayaOkawa, @kento_nishi, @wattenberg, @Hidenori8Tanaka 13/n
TL;DR: Given sufficient context, LLMs can suddenly shift from their concept representations to 'in-context representations' that align with the task structure! w/ @a_jy_l @EkdeepL @YongyiYang7 @MayaOkawa @kento_nishi @wattenberg @Hidenori8Tanaka Paper: arxiv.org/abs/2501.00070 2/n