ICML 2026

Papers presented by Jessica Chudnovsky at ICML 2026

Papers
Consensus is Not Verification: Why Crowd Wisdom Strategies Fail for LLM Truthfulness
Yegor Denisov-Blanch*, Joshua Kazdan*, Jessica Chudnovsky, Rylan Schaeffer, Sheng Guan, Soji Adeshina, Sanmi Koyejo
Thu, Jul 9, 2026 · 5:00 PM – 6:45 PM KST · Hall A #1614
arXiv Tweeprint
Scale Dependent Data Duplication
Joshua Kazdan*, Noam Levi*, Rylan Schaeffer, Jessica Chudnovsky, Abhay Puri, Bo He, Mehmet Donmez, Sanmi Koyejo, David Donoho
Presentation location / time TBD
Accepted at:
  • The Impact of Memorization on Trustworthy Foundation Models (MemFM)
  • Structured Probabilistic Inference & Generative Modeling (SPIGM)
  • Foundations of Deep Generative Models: Understanding Memorization, Generalization, and Reasoning (FoGen)
  • Combining Theory and Benchmarks: Towards a Virtuous Cycle to Understand and Guarantee Foundation Model Performance (CTB)
arXiv Tweeprint
Internal Data Repetition Destroys Language Models
Jessica Chudnovsky*, Joshua Kazdan*, Noam Levi*, Rylan Schaeffer, Yegor Denisov-Blanch, Bo He, Mehmet Donmez, Sanmi Koyejo, David Donoho
Oral at FoGen: Foundations of Deep Generative Models workshop · Fri, Jul 10, 2026 · 11:15 – 11:30 AM KST · Room 318
FoGen Best Paper Runner-up Award · Award ceremony: 4:50 – 5:00 PM KST, July 10 · Room 318
Also accepted at:
  • The Second Workshop on the Impact of Memorization on Trustworthy Foundation Models at ICML
  • Structured Probabilistic Inference & Generative Modeling
  • Combining Theory and Benchmarks: Towards a Virtuous Cycle to Understand and Guarantee Foundation Model Performance
  • High-dimensional Learning Dynamics
arXiv Tweeprint
Seoul — Line Friends store Seoul — Cheonggyecheon area

Images taken from my last time in Seoul to denote my enthusiasm for ICML 2026 taking place in Seoul.

Twitter LinkedIn jchud@cs.stanford.edu