The I Can’t Believe It’s Not Better workshop has been accepted at ICLR 2026. This edition focuses on Where Large Language Models Need to Improve, highlighting limitations, negative results, and careful analyses that are often overlooked but critical for real progress. Call for papers coming soon.

Deadline extension: 1 week.
Please submit your talk proposals by October 16 (anywhere on Earth).

bsky.app/profile/undo...

In addition, the programme and the recorded talks from the previous edition are available, if you would like to get a sense of the range of the conference:
• Programme: undonecs.sciencesconf.org/page/program...
• Recorded talks: mediaserver.univ-nantes.fr/channels/#un...

Picture credit: “Dusk in Luxembourg Grund” <https://www.flickr.com/photos/kewl/28998138111> by Tristan Schmurr (modified), CC BY 2.0

• Luxembourg, 23-25th March 2026 (hybrid)
• Calling for short talk proposals (1-3 pages abstracts)
• Submission deadline: October 9th 2025 (anywhere on Earth)
• Post-proceedings model: we will send a call for full papers after the conference

More info at www.undonecs.org/2026/

Announcing Undone CS 2026:
2nd conference on Undone science in Computer science

A conference in computer science to pause and reflect on the epistemological and ethical dimensions of the field, through the concept of undone science

Full call for presentations at www.undonecs.org/2026/cfp.html

A call for full papers is then issued after the conference for publication in a journal.

More info: www.undonecs.org

We follow a post-proceedings model: short talk proposals (1-3 pages) are selected by a programme committee for presentation at the conference. We accept exploratory contributions that would benefit from discussion at the conference prior to their development into full papers. (6/7)

Organised by computer scientists, we aim to attract colleagues from across the field, as well as philosophers and historians of science, social scientists, and other scholars working on computer science. (5/7)

for instance related to limitations of available methodologies, blind spots of dominant paradigms, institutional and industrial biases, lack of social representation, or other factors. (4/7)

We welcome discussions about the production and dissemination of knowledge in computer science (whether in general or in specific subdomains, whether past or present); (3/7)

We propose the concept of undone science as a line of inquiry: the notion that areas of research may remain incomplete, overlooked, or unfunded despite being found worthy of exploration by some—and the exploration into the causes of these situations. (2/7)

Introduction👋

*Conferences on undone science in computer science*

Our goal is to provide a space to pause and reflect on the epistemological and ethical aspects of computer science. (1/7)