LLMs4PCG 2025
The LLMs4PCG Competition
Character-like Level Generation for Science Birds
What is the LLMs4PCG Competition
The LLMs4PCG Competition builds upon the challenging and exciting spirit of the first and second ChatGPT4PCG competitions while introducing significant changes. This edition expands the competition beyond ChatGPT to include a variety of open-source large language models (LLMs). Participants are challenged to develop prompts or prompt engineering techniques that can construct stable Science Birds levels resembling uppercase English characters using these diverse LLMs.
We’ve enhanced the competition platform to support a wide range of open-source LLMs, allowing participants to explore and compare the capabilities of different models. Submissions can now be in the form of Python programs, enabling more complex prompt engineering techniques that utilize conditions, iterations, and advanced logic. This opens up possibilities for developing novel approaches and potentially creating new prompt engineering methods tailored to specific LLMs.
The competition welcomes participants of all skill levels, from those who simply modify existing prompts to those who develop sophisticated prompt engineering techniques. All submitted programs will be evaluated for qualification according to the competition rules and used to generate levels for each target English uppercase character across multiple LLMs. The generated levels will be assessed for stability using our Science Birds Evaluator and checked for similarity to target characters using an upgraded Vision Transformer (ViT) classifier. We retain the diversity metric introduced in ChatGPT4PCG 2, challenging participants to create prompts that generate stable, similar, and diverse levels across different open-source LLMs. This multi-model approach provides a unique opportunity to explore the strengths and limitations of various LLMs in the context of procedural content generation.
Submission deadline
Midterm: 30 May 2025 (23:59 JST)
Final: 30 July 2025 (23:59 JST)
Midterm submission is optional, although we recommend it. Any team that submits during the mid-term submission will be notified of the preliminary results. However, all teams, whether they submit during the midterm or not, must submit during the final submission period. Only submissions during the final submission period will be considered for the final ranking.
Other Information
Keywords: Prompt engineering, Procedural content generation, Level generation, Conversational agent, Large language model, Angry Birds, Science Birds
Programming languages: Not mandatory: Participants may choose to modify only a prompt in a provided example. However, possessing general programming knowledge can be beneficial. Examples of prompt engineering programs are provided in Python.
Complexity: Low-Medium
Barrier of entry: Low
Contact Us
If you have any questions, please contact us at this email address.
Email address: chatgpt4pcg@gmail.com