E-Values NeurIPS 2026 Submit

NeurIPS 2026 Workshop

E-Values: From Statistics to ML

Bringing together statistics, machine learning, game-theoretic probability, and sequential analysis to advance e-values in modern ML systems.

Recent years have seen a paradigm shift in hypothesis testing and uncertainty quantification through e-values: non-negative random variables whose null expectation is bounded by one. They support anytime-valid inference, continuous monitoring, and data-adaptive decision making, making them especially relevant for modern ML systems that are evaluated, audited, and updated sequentially. This workshop will bring these threads together through invited and contributed talks, posters, a panel discussion, and an open problem session.

Call for Papers Workshop Schedule

Topics

From safe testing to deployed ML systems

We welcome polished results as well as early ideas, emerging connections, practical case studies, and open questions that help define the role of e-values in machine learning.

Foundations & Theory

  1. E-values, e-processes, test martingales, and game-theoretic probability
  2. E-values as evidence, including links to p-values and Bayes factors
  3. E-values, Bayesian methods, pseudo-Bayesian methods, and e-posteriors
  4. Multiple testing, FDR, FWER, and empirical-Bayes connections
  5. Compositional e-value guarantees across multi-stage or multi-agent workflows

Sequential & Adaptive Inference

  1. Sequential monitoring, evidence aggregation, and stopping rules with e-values
  2. Anytime-valid confidence intervals, prediction sets, and safe testing
  3. Optional stopping, optional continuation, and continuous monitoring
  4. Conformal prediction, adaptive coverage, and e-based prediction sets
  5. A/B testing, adaptive experimentation, bandits, and online learning regret

ML Applications, Auditing & Case Studies

  1. Uncertainty under prompt variation, adaptivity, and distribution shift
  2. E-values for LLM evaluation, foundation models, and deployed AI systems
  3. Auditing and verification for tool use, external feedback loops, fairness, and privacy
  4. Applications in science, medicine, safety-critical systems, and decision support
  5. Software, computation, and practical case studies with e-values

Invited Speakers

Confirmed invited speakers

The invited program brings together experts spanning e-values, sequential analysis, multiple testing, bandits, economics, conformal prediction, and reliable machine learning.

Schedule

Single-day workshop program (Tentative)

The final date, schedule, and detailed speaker assignments will be updated once NeurIPS scheduling is confirmed.

Morning

Opening Remarks

Invited Talk 1

Invited Talk 2

Poster Session & Coffee Break

Invited Talk 3

Contributed Oral Presentations

Afternoon

Lunch Break

Invited Talk 4

Invited Talk 5

Panel Discussion: The Future of e-Values in ML

Coffee Break

Poster Session & Closing Remarks

Call for Papers

We invite submissions on e-values and modern inference for ML.

Submissions may cover theoretical, algorithmic, or practical aspects of e-values and related ideas. The workshop is non-archival.

Submission Guidelines

  • Short papers up to 4 pages, excluding references and optional appendices.
  • Papers should use the standard NeurIPS 2026 LaTeX template .
  • Submissions are double-blind and must be anonymized.
  • Accepted papers will be presented as posters, with a subset selected for oral talks.

Important Dates

Deadlines

All deadlines are 23:59 Anywhere on Earth (AOE).

Submission deadline August 29, 2026
Notification of acceptance September 29, 2026
Camera-ready due TBD
Workshop date December 12 or 13, 2026 (TBD)

Organizers

Organizing committee

The organizing team (arranged in alphabetical order by last name) brings together expertise across statistics, sequential inference, and machine learning.