Call for Papers
What we are looking for
The I Can't Believe It's Not Better (ICBINB) initiative invites submissions to ICBINB-BIOat NeurIPS 2026 in Sydney, dedicated to negative results and rigorous evidence of where AI systems for biology fails. We welcome researchers and practitioners who can share unexpected challenges in developing, evaluating, and deploying models on biological tasks and data. Sharing these experiences helps the community avoid unproductive paths, clarify failure modes and boundary conditions, and build a culture of transparency.
To this aim, we invite submissions that:
- Showcase and investigate important limitations of current AI systems in biological settings, including pitfalls in common modeling approaches and evaluations of real-world or deployment-relevant applications.
- Attempt promising ideas to overcome common challenges but fall short of expected gains, accompanied by analyses that clarify failure modes and boundary conditions.
We welcome empirical negative results, unexpected model behavior, and careful analyses of distribution shift, confounding, weak supervision, resource constraints, benchmark misalignment, uncertainty, causal learning, and adaptive strategies in biological applications.
Topics
Around these negative results, topics of interest include (but are not limited to):
- Out-of-distribution generalization and domain shift
- Failure under weak or confounded supervision
- Causal mechanisms versus spurious correlation
- Uncertainty, calibration, and decision-aware reliability
- Interpretability and trustworthy biological inference
- Learning with limited data and distribution shift
- Deployment-relevant evaluation beyond benchmarks
- Limits of foundation, multimodal, and agentic models for biology
- Causal intervention and experimental design
Formats
Full papers
Full submissions may use up to eight pages, excluding references and appendices. Submitted papers should contain the following:
- Problem. A clearly specified biological task, data modality, and setting, with assumptions, target metrics, and desired improvements precisely stated.
- Proposed approach. The modeling approach or solution under investigation, including its core mechanism, required preconditions, and the hypotheses under which it is expected to work.
- Observed outcome. A concise description of the negative, null, or unexpected outcome, with quantitative evidence where possible (metrics, uncertainty, compute budget, data/seed details).
- Reason for failure. An investigation of why the approach did not work as hoped, for example dataset artifacts, modeling choices, confounding, shortcut cues, distribution shift, weak supervision, evaluation mismatches, or deployment constraints, supported by diagnostics or ablations and ending with boundary conditions and actionable takeaways.
Tiny papers
Tiny papers may use up to four pages of main text, with the same formatting instructions as full submissions. They are not required to contain all four elements above, but must state the problem and provide evidence of at least one negative or unexpected outcome, even when the full causal analysis is still developing.
Formatting and review
- Use the workshop LaTeX template: download style files.
- Submit on OpenReview. Submissions are double-blind; linked material must also preserve anonymity.
- Authors may include unlimited appendices, but reviewers are not required to read.
- Disclose any use of large language models (LLMs) with a short paragraph describing their role. We also recommend ethics and reproducibility statements; these do not count toward the page limit. See the NeurIPS Code of Ethics for guidance.
- Accepted papers will appear on OpenReview, and the workshop remains non-archival.
- Concurrent submissions are welcome when they comply with the target venue’s policy.
- We welcome first-time authors. The workshop will be held in person.
Eligibility
Submissions are considered only if the work has not been accepted for publication in previous conference proceedings. Work that is concurrently under review for the NeurIPS 2026 main track is welcome.
Evaluation criteria
- Clarity of the problem and claims.
- Technical rigour and reproducibility.
- Faithfulness to the biological setting.
- Depth of failure analysis.
- Quality of empirical documentation.
- Novelty and significance of the insights.
- Quality of discussion of limitations.
Reviewers may nominate papers for spotlight or contributed talks, as well as the Entropic Award (most surprising negative result) and the Didactic Award (most well-explained and pedagogical paper).
Tentative dates
- Submission
August 29, 2026September 2, 2026- Review
August 29 - September 21, 2026September 2, 2026 - September 21, 2026- Notification
- September 29, 2026
- Camera-ready & poster
- October 20, 2026
- In-person workshop
- December 11 or 12, 2026
All deadlines are 11:59 p.m. Anywhere on Earth.
Updates will be posted on this website. For questions, please contact icbinbbio@gmail.com. We look forward to your submission.