Machine Learning/CORE Unranked

NeurIPS 2026

NeurIPS 2026 Workshop on AI for Verifiable Coding

Dec 12 2026Atlanta, Georgia, United StatesOfficial workshop site Site reachable

The NeurIPS 2026 VeriCodeGen workshop focuses on AI for Verifiable Coding, bringing together researchers in LLMs, formal methods, and programming languages to develop human-aligned agents that collaborate with proof assistants, solvers, and static analyzers to generate code and proofs with machine-checkable guarantees. The workshop emphasizes neuro-symbolic architectures, autoformalization, proof engineering, and evaluation protocols that prioritize end-to-end verification over traditional metrics like pass@k.

Official CFP Back to deadlines Verified September 9, 2026

Key deadlines

Verified September 9, 2026

Abstract registration

September 12, 2026

AoE

Full paper

September 14, 2026

AoE

Workshop timeline

Submission and decisions

Abstract registrationKey deadline

September 12, 2026 · AoE

Full paperKey deadline

September 14, 2026 · AoE

Paper fit

Contribution paths

A strong submission should clearly identify its contribution and evaluate it appropriately.

Research Paper Track

Submissions must be 4–9 pages (excluding references), use the official LaTeX template, and be anonymized for double-blind review. Technical appendices have no page limit but are included in the 50MB file size cap. Supplementary material up to 100MB is allowed and must be anonymized.

Competition Track

Participants submit systems for the Lean Refactor Arena, with two tracks: closed-source LLMs (≤$3 API spend per problem) and open-source LLMs (≤4×80GB A100, ≤48h). Entries require a 4–9 page tech report using a separate template, submitted on OpenReview. Reports are single-blind and must include approach, models, budget, and reproduction details.

Research areas in scope

Neuro-symbolic architectures

LLM agents integrated with proof assistants, model checkers, solvers, static analyzers, and compiler feedbackverifier-guided training and inference

Specification generation and quality

autoformalizationspecification miningambiguity and inconsistency detectionproperty-based validationhuman-guided refinement

Proof search and engineering

premise selectiontactic learningproof repairmaintenancescalabilitycompatibility across theorem-proving systems

Program verification and repair

deductive verificationmodel checkingsymbolic executionabstract interpretationtestingdebuggingcertified repairequivalence checkingbehavior preservation

Verification applications

securitysmart contractsmemory safetysupply chainsaccess-control policiesnetwork configurationsinfrastructure-as-codedistributed protocols

Verification of and for AI

formal validation of generated outputsruntime monitoring of agentsverified guardrailsconstrained tool use

Solver and trusted infrastructure

scalable SAT/SMT solvinglearned heuristicsproof certificatesindependent checkingtrusted computing basesverifier validation

Human-centered verification

mixed-initiative interfacesexplanationstrust calibrationaccessibility for developers without formal-methods expertise

Benchmarks and deployment

datasets and metrics for end-to-end guaranteesscalability studiesreproducibilitynegative resultsindustrial experience

Policies worth checking twice

Submissions must be anonymized for double-blind review; no identifying information in main text, appendices, or supplementary material.
Authors must not use 'our previous work' in citations; anonymize self-citations as 'Smith et al. [1]'.
Abstract submission is mandatory and must be substantive; placeholder or substantially changed abstracts risk desk rejection.
Author list cannot be changed after the abstract deadline; additions or removals require written justification and program chair approval.
LLM use is permitted but must be fully disclosed in methodology; authors are responsible for all content, including text and figures.
Supplementary material (code, data) must be anonymized and is subject to reviewer discretion.
No acknowledgments allowed in submissions.
Submissions must be original, unpublished work in English.

Official sources

Compiled from the official call for papers. The organizers’ pages remain authoritative.

Official workshop site Guidance for VeriCodeGen CFP at NeurIPS 2026

Last verified September 9, 2026