Call for Open Problems

COLT 2026 will include an Open Problems session. As in previous years, selected open problems will also appear in the conference proceedings.

This year, the session will include two types of contributions:

  • Invited open problems, solicited from a small number of researchers, and
  • Regular open problems, submitted through this call.

Because part of the session is reserved for invited contributions, fewer regular open problems will be selected for presentation than in previous years. Accordingly, the regular open problems track is expected to be more competitive than usual.

We warmly encourage submissions of beautiful, important, and well-formulated open problems in learning theory. At the same time, we would like to emphasize a basic standard that every author should ask themselves before submitting: If this open problem were solved, would the resulting paper be a clear accept to COLT? If the answer is not a clear and unambiguous yes, then please reconsider whether the problem is a good fit for this session.

A strong open problem submission should do more than merely state a question. It should explain why the problem matters, place it in context, and make a convincing case that it is both natural and significant for the COLT community. In particular, submissions should:

  • clearly state the open problem,
  • explain its motivation and significance,
  • provide relevant background and prior work,
  • describe any known partial results, related conjectures, or barriers,
  • and be self-contained and accessible to a broad theoretical learning audience.

We welcome both newly formulated open problems and previously stated ones, provided that the submission offers a compelling perspective and a strong case for the problem’s relevance to COLT.

Submission guidelines

  • Deadline: May 25, 2026 (Anywhere on Earth)
  • Format: submissions should use the standard COLT 2026 style files
  • Length: up to 4 pages, excluding references
  • Title: the title should begin with “Open Problem:”
  • Anonymity: submissions are not anonymous; please include author names

Accepted submissions will appear in the COLT 2026 proceedings, and final versions must therefore also use the COLT 2026 style files.