Submission Link: https://icse2027-llm4code.hotcrp.com/

1. Overview

LLM4Code brings together researchers and practitioners who develop, evaluate, and deploy large language models and agents for software engineering tasks. The workshop provides a forum for sharing new methods, systems, datasets, and empirical findings; discussing emerging ideas and limitations; and identifying open challenges in building effective, reliable, efficient, and trustworthy LLMs and agents.

We welcome contributions from academia and industry, including original research, early results, experience reports, replication studies, negative findings, and well-argued positions. Our goal is to advance both the capabilities of LLM-based software engineering and our understanding of when, how, and under what conditions these systems are useful.

2. Paper Types and Page Limits

This workshop accepts research papers, position papers, and extended abstracts:

  • Research papers (4–8 pages, including references): substantial contributions describing new approaches, tools, datasets, or empirical studies.
  • Position papers (1–4 pages, including references): well-argued perspectives, emerging ideas, or research agendas.
  • Extended abstracts / short papers (1–5 pages, including references): concise reports of early results, demonstrations, experience, or work in progress.

3. Submission Requirements

  • Manuscripts must be written in English and submitted as PDF files through the workshop’s HotCRP site.
  • Use the official IEEE conference proceedings template. LaTeX authors should use the following document class, without the compsoc or compsocconf options:
\documentclass[10pt,conference]{IEEEtran}
  • Submissions must be original, must not have been published or accepted elsewhere, and must not be under review at another venue when submitted.
  • The proposed review model is double-anonymous: author identities should be omitted, and self-citations should be written in the third person. The organizers should confirm this policy before opening submissions.
  • Alterations to the required template, spacing, or font sizes may result in desk rejection. Extra pages cannot be purchased.

4. Publication and Presentation

  • Accepted archival papers will appear in the IEEE ICSE 2027 workshop proceedings.
  • The official publication date of the workshop proceedings is the date the proceedings are made available by IEEE. This date may be up to two weeks prior to the first day of ICSE 2027. The official publication date affects the deadline for any patent filings related to published work.
  • Non-archival option: authors may request that an accepted paper be posted only on the workshop website and excluded from the formal proceedings.
  • At least one author of every accepted paper must register for the workshop and present the work.

5. Best Paper Awards

🏆 We will announce best paper awards (up to 10% of accepted papers) during the workshop!

6. Topics of Interest

Topics of interest include, but are not limited to, the following:

  • LLM agents and developer workflows
    • Autonomous and interactive coding agents
    • Repository-scale and long-horizon software-engineering tasks
    • Tool use, planning, memory, context management, and multi-agent coordination
    • Human–agent collaboration, developer experience, and mixed-initiative systems
    • Efficiency, robustness, observability, and operational reliability
  • LLM applications to software-engineering tasks
    • Code generation, completion, and synthesis
    • Fuzzing, test generation, GUI testing, and mobile-app testing
    • Fault localization, debugging, and program repair
    • Vulnerability detection, secure coding, and software assurance
    • Program analysis, code comprehension, and reverse engineering
    • Refactoring, migration, evolution, maintenance, and documentation
  • Datasets, evaluation, and empirical evidence
    • Datasets for pre-training, post-training, and evaluation
    • Automated data generation, augmentation, filtering, and curation
    • Benchmarks, metrics, and evaluation methodology for models and agents
    • Empirical studies of effectiveness, cost, reliability, and developer impact
    • Reproducibility, contamination, leakage, and validity of automated evaluators
  • Model design, training, and optimization
    • Model architectures and pre-training objectives
    • Prompting, context engineering, retrieval, and adaptation
    • Alignment, fine-tuning, preference optimization, and distillation
    • Quantization, inference optimization, and resource-efficient deployment
  • Responsible and trustworthy LLM4Code
    • Security, privacy, safety, and misuse resistance
    • Verification, explainability, uncertainty, and hallucination mitigation
    • Bias, licensing, intellectual property, provenance, and data governance