Call for participation:
SemEval-2026 Task 13: Detecting Machine-Generated Code
We announce an exciting and challenging new SemEval task!
With the rapid growth of large language models for code generation, it is becoming increasingly difficult to distinguish between code written by humans and code produced by AI systems. This raises serious concerns for academic integrity, hiring evaluations, and software security.
To address this, we are introducing SemEval-2026 Task 13: Detecting Machine-Generated Code with Multiple Programming Languages, Generators, and Application Scenarios. It is focused on building systems that can tell whether code was authored by humans or by LLMs.
The task includes three subtasks:
Subtask A: Binary Classification (Human vs. Machine)
https://www.kaggle.com/t/99673e23fe8546cf9a07a40f36f2cc7e
A real-world challenge with strong out-of-distribution shifts in the test set. Participants will need to design robust models capable of adapting to unseen programming languages and contexts: for instance, determining whether a system trained on competitive programming data can generalize to identifying LLM-authore code in GitHub repositories.
Subtask B: Multiclass LLM Authorship Identification
https://www.kaggle.com/t/65af9e22be6d43d884cfd6e41cad3ee4
Participants must identify which specific LLM produced a given code snippet. This subtask aims to explore how stylistic and structural differences across models can support source attribution.
Subtask C: Four-Way Hybrid Authorship Classification
https://www.kaggle.com/t/005ab8234f27424aa096b7c00a073722
A 4-class classification task task. We ask participants to build a system which can not only identify fully haman or LLM-generated codes, but also to find codes with hybrid authorship (human written and then LLM adjusted) and more seriously, codes that were generated by LLMs in attempt to fool detectors (either by prompting or by RLHF fine-tuning).
If you're interested in tackling one of the most relevant and high-impact challenges in code forensics and AI transparency, explore our GitHub repository:
https://github.com/mbzuai-nlp/SemEval-2026-Task13