Adversarial Coevolutionary Illumination with Generational Adversarial MAP-Elites
Unlike traditional optimization algorithms focusing on finding a single optimal solution, Quality-Diversity (QD) algorithms illuminate a search space by finding high-performing solutions that cover a specified behavior space. However, tackling adversarial problems is more challenging due to the behavioral interdependence between opposing sides. Most applications of QD algorithms to these problems evolve only one side, thus reducing illumination coverage. In this paper, we propose a new QD algorithm, Generational Adversarial MAP-Elites (GAME), which coevolves solutions by alternating sides through a sequence of generations. Combining GAME with vision embedding models enables the algorithm to directly work from videos of behaviors instead of handcrafted descriptors. Some key findings are that (1) emerging evolutionary dynamics sometimes resemble an arms race, (2) starting each generation from scratch increases open-endedness, and (3) keeping neutral mutations preserves stepping stones that seem necessary to reach the highest performance. In conclusion, the results demonstrate that GAME can successfully illuminate an adversarial multi-agent game, opening up interesting future directions in understanding the emergence of open-ended coevolution.
Code (1)
Similar Papers 제목 키워드 기반
Tournament Informed Adversarial Quality Diversity
Quality diversity (QD) is a branch of evolutionary computation that seeks high-quality and behaviorally diverse solutions to a problem. While adversarial problems are common, classical QD cannot be easily applied to them…
Multi-population GAN Training: Analyzing Co-Evolutionary Algorithms
Generative adversarial networks (GANs) are powerful generative models but remain challenging to train due to pathologies suchas mode collapse and instability. Recent research has explored co-evolutionary approaches, in w…
DiversityEvolutionary AlgorithmsSelf-Referential Quality Diversity Through Differential Map-Elites
Differential MAP-Elites is a novel algorithm that combines the illumination capacity of CVT-MAP-Elites with the continuous-space optimization capacity of Differential Evolution. The algorithm is motivated by observations…
DiversityCoevolutionary Neural Population Models
We present a method for using neural networks to model evolutionary population dynamics, and draw parallels to recent deep learning advancements in which adversarially-trained neural networks engage in coevolutionary int…
Cross-Generational Transfer of Adversarial Attacks Reveals Non-Monotonic Safety Alignment in LLMs
Safety alignment in LLMs does not improve monotonically across model generations. Studying four generations of Google's Gemma family (7B-31B) with quality-diversity evolution (MAP-Elites) as an automated red-teaming prob…