@lexin_zhou Profile picture

Lexin Zhou

@lexin_zhou

Incoming intern at @MSFTResearch | CS Alumnus at Cambridge | Work on Evaluation, Social Computing, Safety and NLP | Looking for 25Fall PhD position

Pinned

1/ New paper @Nature! Discrepancy between human expectations of task difficulty and LLM errors harms reliability. In 2022, Ilya Sutskever @ilyasut predicted: "perhaps over time that discrepancy will diminish" (youtu.be/W-F7chPE9nU, min 61-64). We show this is *not* the case!

Tweet Image 1

Lexin Zhou Reposted

Interested in #FactChecking ? Come today to our #EMNLP2024 workshop @FEVERworkshop to hear invited talks by the excellent @kristahopsalong @radamihalcea @PCunliffeJones and chris.bregler.com , and learn about our shared task on the AVERITEC dataset fever.ai/workshop.html


Lexin Zhou Reposted

Youmna @YoumnaH presenting An LLM Feature-based Framework for Dialogue Constructiveness Assessment arxiv.org/abs/2406.14760 (led by @lexin_zhou ) on Tuesday


Looking for a Postdoc position and wanna make an impact on developing new robust evaluation frameworks for general-purpose AI systems like LLMs? Please apply to this amazing position at TU Valencia working with Prof. José Hernández-Orallo (josephorallo.webs.upv.es)! He's a legend,…


Lexin Zhou Reposted

A snippet of the work we did over the summer during my internship with MSR FATE Montreal—we highlight the need for more work investigating the impacts of anthropomorphic AI systems, which is critical to understanding the impact of genAI systems: arxiv.org/pdf/2410.08526

Tweet Image 1

Read the arXiv version before. Definitely thought-provoking and recommend it. Congrats @katie_m_collins!

What does it take to build AI systems that meet our expectations and complement our limitations? Our Perspective on thought partners which engage deeply with computational cognitive science is now out in @NatureHumBehav ! nature.com/articles/s4156…



Lexin Zhou Reposted

What happens as chatbots get "bigger and better"? They become more likely to generate wrong answers than to admit ignorance (and sadly people aren't good at spotting the bad answers). Read more in this @nature paper nature.com/articles/s4158…


United States Trends
Loading...

Something went wrong.


Something went wrong.