Full papers
Submissions must follow the exact same paper type and format as NeurIPS 2026 main track submissions, with double-blind peer review.
Machine Learning/CORE Unranked
FLLMPT NeurIPS 2026
NeurIPS 2026 Workshop on Foundations of LLM Post-Training in Changing Environments
FLLMPT 2026 is a NeurIPS workshop focused on developing theoretical and statistical foundations for large language model (LLM) post-training in non-stationary, evolving environments. It addresses challenges such as task drift, feedback loops, and robustness in adaptive post-training, bringing together researchers from machine learning theory, reinforcement learning, and AI safety.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Workshop Topics
Preference feedback as dataStatistical modeling of preference signals, including identifiability questions, heterogeneous annotators, varying feedback strength, and principled treatments of noise and misspecificationRobustness and valid inferenceConditions under which post-training updates are robust to modeling choices, data collection effects, or feedback errors, as well as methods for uncertainty quantification, calibration, and principled evaluationAdaptive data collection and feedback loopsEffects of sequential or adaptive data collection on post-training, including selection bias, feedback loops, active query design, and evolving standards or evaluation criteriaAdaptation mechanisms and limitsTheoretical understanding of post-training mechanisms such as parameter-efficient adaptation, modular updates, or selective fine-tuning, including limits of adaptation, trade-offs with capability preservation, and safety-relevant failure modes
Policies worth checking twice
- All submissions undergo double-blind peer review.
- Authors must anonymize submissions and avoid self-identifying references.
- FLLMPT 2026 is a non-archival workshop.
- Submissions under review at other venues or previously published work are welcome, provided they are updated or recontextualized.
- Prior publication must be disclosed in the submission form.
- Accepted papers will be published on OpenReview after the camera-ready revision period but do not prevent future submission to archival venues.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026