Large language models compete and learn by creating and solving tasks
Environment Duels: A Microcosm of LLM Co-Evolution
Computation and LanguageArtificial Intelligence
Summary
Large language models (LLMs) learn by interacting with content made by earlier versions or other models. This paper presents Environment Duels, a setup where models both create puzzles with hints and try to solve them using those hints. The study finds that being good at solving tasks and creating hard tasks are different skills among these models. Smaller models tend to write the toughest puzzles and best hints by learning from others, while stronger solvers aren’t always the best creators. This method offers a way to observe how models teach and challenge each other in a controlled setting.
What this means in practice
- •For ai developers: Design training systems where language models improve by both composing and solving interactive challenges collaboratively.
- •For software engineers: Create evaluation environments that test AI models’ reasoning and adaptability through puzzles with guided hints.
Authors
Shibo Jin, Zhiqiu Xu, Yiling Zou, Shangyu Gong, Mayur Naik
Abstract
Through a tangle of real-world interactions, from synthetic training pipelines to data shared on the open web, frontier LLMs absorb what their predecessors and competitors create. How well does each model play its role in this ecosystem, as a creator and as a solver? We introduce Environment Duels, a dynamic agentic benchmark that evaluates both capabilities on interactive environments. Each model writes environments, each with a privileged hint, aiming for tasks that capable solvers fail without the hint and pass with it. Across nine frontier LLMs, solving and authoring prove to be distinct skills: the strongest solvers are not the strongest authors, while smaller models learn more from how others played and write the hardest environments and the most useful hints. Beyond the scores, the pairwise interactions form a microcosm of how models affect one another. Head-to-head outcomes are transitive and so yield a stable ranking; independent authors converge on the same designs; and in a reinforcement-learning study with Qwen3.8-27B, the learner's hint lift on an author's environments tracks how much training on those environments improves it on held-out reasoning and agentic benchmarks, whether the environments are its own or another model's. Each of these patterns has a counterpart in how models shape one another outside the duel. By staging these interactions under controlled conditions, Environment Duels offers a window onto how LLMs teach, challenge and learn from one another. Code is available at https://github.com/LLM-Duels/EnvDuels.