Procedural Generation of Algorithm Discovery Tasks in Machine Learning

Automating the development of machine learning algorithms has the potential to unlock new breakthroughs. However, our ability to improve and evaluate algorithm discovery systems has thus far been limited by existing task suites. They suffer from many issues, such as: poor evaluation methodologies; data contamination; and containing saturated or very similar problems. Here, we introduce DiscoGen, a procedural generator of algorithm discovery tasks for machine learning, such as developing optimisers for reinforcement learning or loss functions for image classification. Motivated by the success of procedural generation in reinforcement learning, DiscoGen spans millions of tasks of varying difficulty and complexity from a range of machine learning fields. These tasks are specified by a small number of configuration parameters and can be used to optimise algorithm discovery agents (ADAs). We present DiscoBench, a benchmark consisting of a fixed, small subset of DiscoGen tasks for principled evaluation of ADAs. Finally, we propose a number of ambitious, impactful research directions enabled by DiscoGen, in addition to experiments demonstrating its use for prompt optimisation of an ADA. DiscoGen is released open-source at https://github.com/AlexGoldie/discogen.

Guiding Evolution of Artificial Life Using Vision-Language Models

Foundation models (FMs) have recently opened up new frontiers in the field of artificial life (ALife) by providing powerful tools to automate search through ALife simulations. Previous work aligns ALife simulations with natural language target prompts using vision-language models (VLMs). We build on Automated Search for Artificial Life (ASAL) by introducing ASAL++, a method for open-ended-like search guided by multimodal FMs. We use a second FM to propose new evolutionary targets based on a simulation's visual history. This induces an evolutionary trajectory with increasingly complex targets. We explore two strategies: (1) evolving a simulation to match a single new prompt at each iteration (Evolved Supervised Targets: EST) and (2) evolving a simulation to match the entire sequence of generated prompts (Evolved Temporal Targets: ETT). We test our method empirically in the Lenia substrate using Gemma-3 to propose evolutionary targets, and show that EST promotes greater visual novelty, while ETT fosters more coherent and interpretable evolutionary sequences. Our results suggest that ASAL++ points towards new directions for FM-driven ALife discovery with open-ended characteristics.

Welfare Diplomacy: Benchmarking Language Model Cooperation

The growing capabilities and increasingly widespread deployment of AI systems necessitate robust benchmarks for measuring their cooperative capabilities. Unfortunately, most multi-agent benchmarks are either zero-sum or purely cooperative, providing limited opportunities for such measurements. We introduce a general-sum variant of the zero-sum board game Diplomacy -- called Welfare Diplomacy -- in which players must balance investing in military conquest and domestic welfare. We argue that Welfare Diplomacy facilitates both a clearer assessment of and stronger training incentives for cooperative capabilities. Our contributions are: (1) proposing the Welfare Diplomacy rules and implementing them via an open-source Diplomacy engine; (2) constructing baseline agents using zero-shot prompted language models; and (3) conducting experiments where we find that baselines using state-of-the-art models attain high social welfare but are exploitable. Our work aims to promote societal safety by aiding researchers in developing and assessing multi-agent AI systems. Code to evaluate Welfare Diplomacy and reproduce our experiments is available at https://github.com/mukobi/welfare-diplomacy.