SignalTech/AI
LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets
High-priority alert based on source significance and keyword relevance.
Context
What happened: Researchers introduced LiveMACE, a process-aware benchmark that evaluates persistent LLM agents in live cryptocurrency and US-equity markets using decision traces rather than outcomes alone.
Key numbers: Five frontier LLMs were evaluated over 30 days of live trading across four capability configurations: tool use, persistent memory, rule following,
Why it matters: The work reports an outcome-capability gap, where higher returns do not consistently imply stronger capabilities and similar returns can arise from different patterns of mechanism use. The framework is open source.
What changed
Researchers introduced LiveMACE, a process-aware benchmark that evaluates persistent LLM agents in live cryptocurrency and US-equity markets using decision traces rather than outcomes alone.
Why it matters
The work reports an outcome-capability gap, where higher returns do not consistently imply stronger capabilities and similar returns can arise from different patterns of mechanism use. The framework is open source.