The ACM Transactions on Evolutionary Learning and Optimization publishes high quality original papers in all areas of evolutionary computation and related areas such as evolutionary machine learning, evolutionary reinforcement learning, Bayesian optimization, evolutionary robotics and other metaheuristics. It publishes issues on a quarterly basis.

We welcome papers that make solid contributions to theory, method and applications. Relevant domains include continuous, combinatorial or multi-objective optimization. Applications of interest include but are not limited to logistics, scheduling, healthcare, games, robotics, software engineering, feature selection, clustering as well as the open-ended evolution of complex systems.

We are particularly interested in papers at the intersection of optimization and machine learning, such as the use of evolutionary optimization for tuning and configuring machine learning algorithms, machine learning to support and configure evolutionary optimization, and hybrids of evolutionary algorithms with other optimization and machine learning techniques. ACM TELO encourages reproducibility.

Special Issue on Integrating Evolutionary Algorithms and Large Language Models
截稿日期: 2024-10-01

Guest Editors:
• Erik Hemberg, Massachusetts Institute of Technology, USA, hembergerik@csail.mit.edu
• Una-May O’Reilly, Massachusetts Institute of Technology, USA, unamay@csail.mit.edu
• Dennis Wilson, ISAE-Supaero, University of Toulouse, France, dennis.wilson@isae.fr

Large language models (LLMs), along with other Foundational Models, have surpassed prior expectations of
Artificial Intelligence and Machine Learning when used for fundamental tasks exemplifying human intelligence
such as language, vision, and code processing. An LLM relies upon very large quantities of training data, extensive
training of a highly complex transformer model, and post-training optimization. It processes natural language text
prompts as input and completes the prompt as output. The prompt is written to describe a task, and the model
algebraically combines the encodings of the tokens of the task with its weights to obtain a prediction of what
completes the token sequence. This prediction impressively frequently turns out to be a valid response to the
task, primarily because of the weight training.

In comparison, Evolutionary Algorithms (EA) are inspired by Neo-Darwinian evolution. They rely on a population,
abstract genotype-phenotype mapping, genetic operators, and adroit representation and fitness function design.
Through another lens, EAs and other metaheuristics conduct black-box search and optimization of a solution-
space.

What then brings these two machine learning approaches together? And, how can their integration lead to
interesting design insights, new results, and even new questions, in Evolutionary Computation? These questions
are at the root of this call for papers.

Topics

Topics include, but are not restricted to, the following list:
• LLM-Assisted Evolution: How can Eas and other metaheuristics effectively utilize LLMs for diverse
representations (code, strings, images, multimodal)? How can LLMs enhance the algorithmic aspects of
exploration of cooperation, competition, and information reuse in EAs?
• LLMs for Evolutionary Operators: Can LLMs be creatively used as a part of or to enhance evolutionary
operators such as recombination and variation?
• Novel EA Hybrids: What new variants hybridizing GA, ES, GP and/or another search heuristic with LLMs are
possible? In what respects are they advantageous?
• Multi-Objective and Open-ended Optimization: Can EAs that integrate LLMs leverage multi-objective and
open-ended optimization in compelling ways, and if so, how?
• Scalability and Search Space Analysis: How do LLM-based search heuristics scale with population size and
problem complexity? How can search spaces in these LLM-based heuristics be effectively characterized and
analyzed in relation to problem difficulty?
• Robustness and Benchmarking: How can vulnerabilities and biases in LLMs be mitigated within evolutionary
approaches? What makes a good benchmark for evaluating the performance of LLMs in an EA? Which types
of problem benefit from the combination of EAs and LLMs, and which benchmarks demonstrate this?

Important Dates
• Open for Submissions: June 1, 2024
• Submissions deadline: October 1, 2024
• First-round review decisions: January 1, 2025
• Deadline for revision submissions: March 1, 2025
• Notification of final decisions: June 1, 2025

Submission Information
Manuscripts should be prepared according to the “Guidelines for Authors” section at
https://dl.acm.org/journal/telo/author-guidelines and submissions should be made through the journal
submission website at https://mc.manuscriptcentral.com/telo by selecting the Manuscript Type “Special Issue on
Integrating Evolutionary Algorithms and Large Language Models”. If you would like to apply for an ACM
Reproducibility Badge, please clearly state this in your Cover Letter. Submission of a manuscript implies that it is
not being submitted for possible publication elsewhere.
For questions and further information, please contact one of the guest editors.