5arXiv cs.CL (Computation and Language)·
The Energy Society: Simulation environment for studying LLM agent cooperation under survival pressure
Researchers introduce The Energy Society, a minimal survival-economy simulation where LLM-based agents spend energy proportional to model size when generating tokens and deactivate if energy reaches zero. The environment is used to compare competitive vs. cooperative incentive structures, finding that cooperative incentives lead to donation behaviors and altered job allocation, while larger models consistently spend more energy than they gain. Ablations show that inter-agent action recommendations support coordination and that memory helps agents calibrate risk, while direct sabotage is rare but subtle self-serving behavior emerges in competitive settings.
Related guides (3)
Related events (8)
Agentopia: Long-term multi-agent life simulation framework for training LLMs on social behavior
Researchers introduce Agentopia, a framework for simulating 10 years of social life across 100 LLM-powered agents, enabling study of emergent social behaviors and long-term personal growth dynamics. The system defines a 'life reward' metric mirroring human well-being and uses it to train LLMs via rejection sampling. Training on simulated social experience yields a +15.6% improvement on downstream role-playing benchmarks, suggesting that synthetic social simulation can generalize to real capability gains.
Agentic LLM collectives proposed as interpretable substrates for Artificial Life research
A preprint from arXiv argues that populations of agentic LLMs — equipped with persistent memory, tools, and autonomous action — can serve as a computational substrate for Artificial Life (ALife) research. The key claim is that because agents communicate in natural language, their collective emergent behaviors are directly interpretable by examining textual traces or querying the agents themselves. The paper extends existing notions of LLM interpretability to multi-agent collectives and surveys recent examples of agentic LLM systems in both controlled and deployed settings. This positions multi-agent LLM systems as a novel lens for studying emergence and complexity while retaining interpretability.
Survey: Agentic Environment Engineering for LLMs — Modeling, Synthesis, Evaluation, and Application
A comprehensive arXiv survey systematically reviews the design and engineering of interactive environments for LLM-based agents, covering the full lifecycle from environment modeling and synthesis to evaluation and application. The paper categorizes environments across eight attributes and eight domains, introduces symbolic and neural synthesis paradigms, and characterizes four pathways for agent-environment co-evolution including memory-centric, orchestration-centric, trajectory-centric, and exploration-centric approaches. It also identifies three paradigms of environment evolution (neural-driven, difficulty-driven, scaling-driven) and proposes future directions such as Environment-as-a-Service and multi-agent environments. This is a reference-organizing contribution for the rapidly growing agent tooling and evaluation space.
SolarChain-Eval: Physics-constrained benchmark for trustworthy AI agents in decentralized energy markets
Researchers introduce SolarChain-Eval, a benchmark for evaluating autonomous economic agents in decentralized energy markets under physical constraints. The benchmark formulates market governance as a Gymnasium-compatible MDP and assesses policies across market utility, physical safety, slippage, fairness, and auditability dimensions. An LLM-based Planner/Auditor layer is incorporated to constrain and log agent actions; experiments reveal a utility-safety trade-off where RL agents improve market utility but can still produce unsafe behavior, and the LLM layer cannot compensate for a misspecified reward function. Code and data are released open-access.
Multi-Agent Fictitious Play (MAFP) applies game-theoretic equilibrium-seeking to LLM decision-making
Researchers propose Multi-Agent Fictitious Play (MAFP), a multi-agent system paradigm that frames LLM-based decision-making as an equilibrium-seeking process borrowed from game theory. Each agent represents a stakeholder stance and iteratively best-responds to the empirical mixture of other agents' past decisions, addressing what the authors call 'stance entanglement' — mutual interdependence among stakeholder decisions that cannot be decomposed into independent subtasks. MAFP is evaluated on competitive strategy tasks and outperforms single-round and multi-round baselines on tournament strength and robustness metrics. The work extends the MAS literature beyond divide-and-conquer execution patterns into interdependent decision scenarios.
AgenticSTS: Bounded-memory testbed for studying long-horizon LLM agent decisions in Slay the Spire 2
Researchers introduce AgenticSTS, a testbed for studying long-horizon LLM agents under a bounded-memory contract where each decision is assembled from typed retrieval rather than appending a raw cross-decision transcript. The system is instantiated in Slay the Spire 2, a stochastic deck-building game requiring hundreds of sequential decisions, chosen because frontier LLMs currently win zero games at the lowest difficulty against a 16% human baseline. Ablation experiments show enabling a strategic skill layer improves win rate from 3/10 to 6/10, though sample sizes are too small for statistical significance. The authors release 298 completed trajectories, memory snapshots, and analysis scripts as a reusable methodology for isolating how explicit memory layers affect agent performance.
Thousand Token Wood: multi-agent economy simulation shipped on a 3B model
A Hugging Face blog post documents a hackathon project called Thousand Token Wood, which implements a multi-agent economic simulation running on a 3-billion-parameter model. The project demonstrates that complex multi-agent coordination can be achieved at small model scales. This is a practical case study in lightweight agentic systems from the HuggingFace Build Small hackathon.
Benchmark and study of LLM agents in deliberative multi-agent collaboration under partial observability
Researchers introduce a scalable benchmark for evaluating LLM agents on cooperative joint decision-making tasks where agents must exchange information under partial and asymmetric observations to reach a shared decision. A systematic evaluation of representative LLMs finds that state-of-the-art models still struggle with complex deliberative collaboration, failing in either information alignment or downstream reasoning even with external mathematical tools. Diagnostic analysis also reveals that deliberation can enable reflection and error correction, sometimes outperforming centralized baselines, offering a nuanced picture of multi-agent LLM capabilities.