The Conference on Language Modeling (COLM) 2026 is being hosted at the Hilton Union Square in San Francisco from October 6th - 9th. We’re excited to share all the work from SAIL that’s being presented, and you’ll find links to papers, videos and blogs below. Feel free to reach out to the contact authors directly to learn more about the work that’s happening at Stanford!

Main Conference

Conversation as Measurement in Clinical Encounters: Observable Phase Structure, Partially Observable Patient State

Authors: Lily Chen, Ted Mau, Michael Gensheimer, Brian Anthony Nuyen, Nancy Jiang, James Zou
Contact: l1ly@stanford.edu
Links: Paper
Keywords: language models, conversation, measurement, clinical encounters

Do Language Models Consistently Encode the Current Year?

Authors: Suze van Adrichem, Aditi Bhaskar, Diyi Yang, Christopher Potts, Jing Huang
Contact: suzeva@stanford.edu
Venue: Main conference (also at the Actionable Interpretability Workshop)
Links: Paper
Keywords: causal interventions, training data cutoff, refreshing language models, temporal alignment

PromptNCE: Conditional Probabilities and PMI Using Only LLMs and Contrastive Estimation Prompts

Authors: Juliette Woodrow, Chris Piech
Contact: jwoodrow@stanford.edu
Links: Paper
Keywords: mutual information, pointwise mutual information, large language models, conditional probability estimation, contrastive estimation, prompt-based estimation

SWE-chat: Coding Agent Interactions From Real Users in the Wild

Authors: Joachim Baumann, Vishakh Padmakumar, Xiang Li, John Yang, Diyi Yang, Sanmi Koyejo
Contact: joachimbaumann@stanford.edu
Award nominations: Oral Presentation at COLM Workshop on Responsibly Enabling Data for Foundation Models
Links: Paper | Website
Keywords: in-the-wild data, vibe coding, coding agents, human-ai interaction

Switching Linear Attention

Authors: Hyun Dong Lee, Xavier Gonzalez, Nicolas Zucchet, E. Kelly Buchanan, Emily B. Fox, Scott W. Linderman
Contact: hdlee@stanford.edu
Links: Paper | Website
Keywords: recurrent architecture, sequence modeling, test-time regression, mixture model

Workshop Papers

Neural Garbage Collection: Learning to Forget while Learning to Reason

Authors: Michael Y. Li, Jubayer Ibn Hamid, Emily Fox, Noah Goodman
Contact: michaelyli@stanford.edu
Workshop: COLM 2026 Workshop CBW
Award nominations: Oral Presentation
Links: Paper
Keywords: rl, kv cache, end-to-end context management

QuasiMoTTo: Quasi-Monte Carlo Test-Time Scaling

Authors: Michael Y. Li, Anthony Zhan, Kanishk Gandhi, Noah Goodman, Emily Fox
Contact: michaelyli@stanford.edu
Workshop: Workshop on Efficient Reasoning
Links: Paper
Keywords: test-time scaling, inference compute

Speculative Self-Distillation enables Efficient Knowledge Internalization

Authors: Shayan Talaei, Agam Bhatia, Arshia Soltani Moakhar, Jonas Hübotter, Amin Saberi, Azalia Mirhoseini
Contact: agam2026@stanford.edu
Workshop: LLA
Links: Paper
Keywords: distillation, post training

Sycophantic Agreement Transfers with Neutral Data via Contrastive Preference Optimization

Authors: Camila Blank, Zhuofan Ying, Christopher Potts, Peter Hase, Jing Huang
Contact: camilab@stanford.edu
Workshop: Actionable Interpretability Workshop
Links: Paper
Keywords: sycophancy, subliminal learning, ai safety and alignment

We look forward to seeing you at COLM 2026!