Null Epoch Season 0 Open: AI Agent Logs from a 10-Day MMO Simulation

A 10-day live MMO simulation dataset containing raw, sanitized interaction logs, metrics, economy transactions, and reasoning traces from 20 autonomous AI agents. It was created by Firespawn Studios, with 17 system agents powered by 8 different LLMs and 3 user-deployed agents. The dataset was last updated on Hugging Face on May 18, 2026.

Use Cases

• Analyze emergent social and economic dynamics based on agent interaction logs and economy transactions.
• Benchmark the reasoning and decision-making capabilities of different LLMs based on the provided reasoning traces.
• Study multi-agent coordination and conflict within a simulated persistent world based on the described metrics and logs.
• Train or fine-tune AI agents for interactive environments using the recorded agent behavior data.

Strengths

• Contains data from a controlled 10-day live simulation, providing a temporal sequence of events.
• Includes logs from 20 distinct autonomous agents, allowing for multi-agent analysis.
• Traces originate from agents powered by a mix of 8 different open-weight and proprietary LLMs, enabling comparative study.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training tasks.
• Description metadata is limited; actual data quality and completeness require manual inspection after download.

Provenance

Source

FirespawnStudios

Collection Method

Collected from a live MMO simulation involving 20 autonomous AI agents.

Time Range

Covers a 10-day simulation period.

Freshness

Last updated 2026-05-18 15:28:07; freshness should be verified.