System 1 agents: a method for local, reliable and fast agentic AI
arXiv — cs.AI · Eleftherios Tzanis, Michail E. Klontzas · Published on Readzer
Summary
The paper presents System 1 agents, which call tools without generating text, using a deterministic compiler to convert tool schemas and requests into multiple-choice questions answered by a one-pass decision model. On 1,952 requests to 157 biomedical tools, the best System 1 agent chose the right tool or correctly declined for 97.0% of requests, outperforming all 18 local LLM-agent settings (best 95.2%). No System 1 call contained invented content, versus 5.5-19.7% of LLM calls, and the framework is open-source at https://github.com/eltzanis/SystemOneAgents.
This is a reviewed summary. Read the complete article at the original source.
Why it matters by industry
- Artificial intelligence
For AI practitioners, this method addresses key limitations of local LLM agents—hallucination and privacy—by eliminating text generation, potentially enabling safer automation of sensitive data analysis in healthcare and other regulated domains.