COLM 2026 · October 9 · San Francisco

DAIH: LLM/VLM Deployment Opportunities and Risks in Healthcare

A full-day workshop on what it actually takes to responsibly deploy language and vision-language models in clinical settings.

Contact: colm.daih2026@gmail.com

Submit Work Call for Papers

Why DAIH

Clinical AI Needs a Deployment Science

Benchmark performance is a poor proxy for real-world readiness. DAIH brings together machine learning researchers, physicians, health-system leaders, and policy researchers to tackle safety, equity, privacy, regulation, workflow fit, and post-deployment monitoring.

Know Whether It Is Safe

Evaluation beyond benchmark accuracy, hallucination and failure characterization, and robustness under temporal and institutional shift.

Know Whether It Is Equitable

Fairness across patient populations, privacy-preserving deployment, legal soundness, and emerging regulatory constraints.

Know Whether It Works

Multimodal clinical applications, human-AI collaboration, operational lessons, and the practical reality of production deployment.

DAIH @ COLM 2026

Call for Papers

We invite non-archival submissions on the responsible deployment of large language and vision-language models in clinical settings.

May 23Submissions open

June 23Submission deadline

July 10Reviews due

July 24Notification

All deadlines are Anywhere on Earth (AoE).

Submit on OpenReview

Topics

We welcome novel work relating to the responsible deployment of LLMs and VLMs in healthcare, including (but not limited to):

Real-world deployment and clinical integration

Multimodal clinical reasoning and decision support
Human–AI collaboration in clinical workflows
Operational integration and workflow design
Clinical documentation, summarization, and chart review
Patient-facing communication and health literacy
Negative results and failure case analyses from real-world deployments
Lessons learned from industry and health system implementations

Safety and reliability

Clinical evaluation beyond benchmark accuracy
Hallucination detection, characterization, and mitigation in medical contexts
Robustness under temporal, institutional, and demographic distribution shift
Failure-mode characterization and safety monitoring
Post-deployment surveillance and error analysis

Equity, privacy, and governance

Fairness and bias auditing across patient populations
Privacy-preserving methods for clinical LLM/VLM deployment
Regulatory compliance and emerging governance frameworks
Accountability and liability in clinical AI systems

Submission Tracks

We solicit non-archival submissions in two categories.

Research Papers

Up to 8 pages excluding references. Technical work on evaluation, safety, robustness, multimodal modeling, fairness, privacy, governance, monitoring, or workflow integration. Accepted papers will be presented as spotlight talks or posters.

Deployment Case Reports

Up to 4 pages excluding references. Practical accounts from clinical, operational, or industry deployments — including work-in-progress, negative results, and lessons learned. Accepted case reports will be presented as posters.

Submission Guidelines

Platform: OpenReview
Format: Papers must use the COLM 2026 LaTeX template.
Review process: Double-blind. Authors should anonymize their submissions accordingly; papers that are not properly anonymized will be desk rejected. Related arXiv papers by the same authors do not break anonymity; if cited, they should be cited in third person.
Conflict of interest: Conflicts will be managed per COLM policy. Conflicted organizers will be recused from assignment, discussion, and decisions for affected papers.
Dual-submission policy: We accept submissions of ongoing unpublished work as well as work currently in submission elsewhere. We also welcome substantial extensions of works previously presented at non-archival venues. However, we do not accept work that has already been published in a journal or included in conference proceedings (including the COLM main conference).
LLM usage policy: Authors must follow COLM 2026 policies on Large Language Model usage.

Program

Workshop Schedule

A full-day program balancing invited talks, contributed spotlights, poster sessions, a structured roundtable lunch, and a deployment-focused panel.