Workshop on Responsibly Enabling Data for Foundation Models @ COLM 2026

About the Workshop

As foundation models scale, available training data sources have rapidly depleted. However, several forms of valuable data artifacts such as medical records, legal, and financial documents are restricted from use in model training due to their sensitive nature. In addition, the strong reasoning capabilities in current generative models have opened the possibility for highly personalizable AI applications but these remain bottlenecked by limited access to high quality user data. Hence, it is of immense value to responsibly unlock these data sources (for example: using data transformation or constrained training paradigms) or to generate synthetic alternatives. In this workshop, we aim to bring together domain experts in data, privacy, model training, and legal policy, to advance the frontier of responsibly leveraging such sensitive data with foundation models.

Topics of interest include (but are not limited to):

Data Transformation: De-identification, Anonymization, Pseudonymization.
Synthetic Data Generation: Controlled Regeneration, Data Diversity.
Novel Training Paradigms: DP, Federated Learning, Architectural Solutions.
Evaluation & Auditing: Privacy attack benchmarks, Utility-Privacy tradeoffs.
Policy: Compliance, New regulations on data sharing.

Call for Papers

We invite long papers with novel research contributions (up to 8 pages long) as well as short papers (up to 4 pages) reflecting preliminary studies or negative results.

Submissions are managed via OpenReview. Accepted papers are non-archival, and concurrent submissions are allowed. Please follow the COLM 2026 template.

Key Dates

All deadlines are 23:59 AoE (anywhere on earth)

~~Submissions open: May 27, 2026~~
Submission deadline: ~~June 23, 2026~~ June 28, 2026 AoE
~~Acceptance Notifications: July 24, 2026~~
Workshops day at COLM: October 9, 2026

Workshop Schedule

08:45 - 09:00	Welcome and Opening Remarks
09:00 - 09:35	Invited Talk 1 "Data as a Binding Resource: How Pre-training Should Change" by Sewon Min (slides)
09:35 - 10:10	Invited Talk 2 "Private Text Generation with LLMs: Mechanisms, Applications, and Audits" by Krishna Pillutla (slides)
10:10 - 10:45	Invited Talk 3 "Overclaiming in Frontier Coding Agents" by Nouha Dziri (slides)
10:45 - 11:00	Coffee Break
11:00 - 12:00	Oral Session SWE-chat: Coding Agent Interactions From Real Users in the Wild Joachim Baumann, Vishakh Padmakumar, Kevin Li, John Yang, Diyi Yang, Sanmi Koyejo ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities? Peihan Liu, Lucas Rosenblatt, Weiwei Kong, Natalia Ponomareva, Gautam Kamath, Rachel Cummings, Roxana Geambasu, Yu Gan, Lillian Tsai, Alex Bie Beyond FineWeb2-HQ through Synthetic Refinement Bettina Messmer, Vinko Sabolčec, Martin Jaggi Sort, Partition, Randomize: Optimal Binary Hypothesis Testing under Local Differential Privacy Elena Ghazi, Jawad Nasser, Flavio Calmon, Ibrahim Issa
12:00 - 13:30	Lunch
13:30 - 14:05	Invited Talk 4 "The Legal Life of a Dataset: What Makes Training Data Usable?" by Janel Thamkul (slides)
14:05 - 14:40	Invited Talk 5 "Creating RL Environments for Responsible Frontier Model Post-Training" by Alex Dimakis (slides)
14:40 - 15:15	Invited Talk 6 "How Does Post-Training Change the Imprint of Pretraining Data?" by Niloofar Mireshghallah (slides)
15:15 - 15:30	Coffee Break
15:30 - 17:00	Poster Session InfoSynth: Information-Guided Benchmark Synthesis for LLMs Ishir Garg, Neel Kolhe, Xuandong Zhao, Dawn Song SWE-chat: Coding Agent Interactions From Real Users in the Wild Joachim Baumann, Vishakh Padmakumar, Kevin Li, John Yang, Diyi Yang, Sanmi Koyejo Inference-Efficient Zero-Shot Differentially Private Text Synthesis William Fang, Mayana Pereira ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities? Peihan Liu, Lucas Rosenblatt, Weiwei Kong, Natalia Ponomareva, Gautam Kamath, Rachel Cummings, Roxana Geambasu, Yu Gan, Lillian Tsai, Alex Bie Natively Unlearnable Large Language Models Gaurav R Ghosal, Pratyush Maini, Aditi Raghunathan Speech Generation Speaker Poisoning: Capability Erasure in Zero-Shot Text-to-Speech Thanathai Lertpetchpun, Thanapat Trachu, Sai Praneeth Karimireddy, Shrikanth Narayanan Beyond FineWeb2-HQ through Synthetic Refinement Bettina Messmer, Vinko Sabolčec, Martin Jaggi Be Kind, Rewrite: Benign Projections via Rewriting Defend Against LLM Data Poisoning Attacks John Halloran, Noopur Bhatt Can we enable speech data with Differential Privacy? Efthymios Tsaprazlis, Vasilis Varsamis, Tiantian Feng, Sai Praneeth Karimireddy, Shrikanth Narayanan Librarian Agents: Anticipatory Filesystem Construction via Reinforcement Learning Anikait Singh, Teresa Zhang, Yoonho Lee, Roshen S Nair, Aviral Kumar, Chelsea Finn Exact Unlearning via Quantized Sufficient Statistics Ami Tavory, Shripad Gade, Tal Sarig, Noam Touitou, Ido Guy Prune, Then Select: Select High-Quality, Important, and Diverse Data Using Training Trajectories Yajiao Liu, Ruoyu Sun Sort, Partition, Randomize: Optimal Binary Hypothesis Testing under Local Differential Privacy Elena Ghazi, Jawad Nasser, Flavio Calmon, Ibrahim Issa Calibrated Candidate Pools for Synthetic EHR Release: Coverage, Concentration, and Privacy Risk Md Nasir, Omid Bazgir, Manu Agrawal, Yang Yang, Vinay R Dandin, Anusua Trivedi, Christine Swisher, Chris Gibbons UDET: A Dataset for Detecting Contamination on Competition Math via Underdetermined Problems Atul Dev, Ishaan Bansal, Shivank Garg, Paras Chopra Pseudonymized but Linkable: Cross-Document Re-identification Risks in LLM-Ready Sensitive Text Corpora Jingxu Xie Forger: An Environment for Agentic Data Generation Vijay Viswanathan, Marshall Mandell, Rishith Prathi, Pang Wei Koh, Graham Neubig, Sherry Wu
17:00 - 17:05	Closing Remarks