[CfP] EVOKE 2026: Eliciting Verifiable, Operational Knowledge from Experts

  EVOKE 2026: Call for Participation

Eliciting Verifiable, Operational Knowledge from Experts


    Full-day workshop at HCOMP and CI 2026September 27, 2026 · Virginia
    Tech, near Washington, DC

Submission deadline: August 15, 2026

Notification: August 22, 2026

Workshop: September 27, 2026

More information: www.evokebench.com <http://www.evokebench.com>


EVOKE invites work on eliciting knowledge into representations that a 
machine can reason over soundly. While large language models offer new 
frontiers in tackling the knowledge acquisition bottleneck, they also 
present new challenges in ensuring the validity and utility of acquired 
knowledge.   When that knowledge is encoded in a formal representation – 
for example, as a constraint set, an optimization model, a logical or 
temporal specification, a program, an ontology, or a probabilistic 
model, its utility becomes measurable. A solver can provide ground 
truth. A verifier can certify a result. A contradiction can be exposed.

The workshop focuses on three connected themes:

 1.

    Expert crowdsourcing for LLM training and post-training, including
    LLM-mediated and mixed-initiative elicitation, clarifying questions,
    tacit knowledge, and the knowledge-acquisition bottleneck.

 2.

    Verified elicitation into formal models, including integration of
    language models with solvers, SAT, SMT, MILP, constraint
    programming, and theorem provers.

 3.

    Elicitation for operations research, including ORBench, eliciting
    constraints and optimization models, and combining operations
    research with language models.

EVOKE will work toward:

  *

    a position paper on verified elicitation; and

  *

    a working group for a verified-elicitation dataset and benchmark.

We welcome technical papers, position papers, demonstrations, 
work-in-progress, and datasets. Accepted contributions will be presented 
as a combination of posters and panel discussion. To submit, email a 
title and one-page abstract to: evokebench@googlegroups.com 
<mailto:evokebench@googlegroups.com>