Automata from Agent Traces

These machines replay held-out data at 0.997 fitness and build in milliseconds.

FSM-state context outperforms Agent Workflow Memory on all ground-truth-matched datasets for next-step prediction.

Per-state behavioral features achieve held-out AUROC up to 0.94 for failure prediction.

Another good paper. Interesting finding on the benefits of the agent harness. DAIR.AI @dair_ai // Automata from agent traces // How much of your agent's behavior comes from the model, and how much from the harness you wrapped around it? New work collapses an entire corpus of agent traces into a single compact finite-state machine. Across twelve public datasets the induced machines run 7 to 43 states, replay held-out data at 0.997 fitness with near-identical topology across splits, and build in milliseconds. FSM-state context beats Agent Workflow Memory on every ground-truth-matched dataset for next-step prediction. Per-state behavioral features reach held-out AUROC up to 0.94 for failure prediction, and an online monitor ranks failing runs above passing ones from a partial trace, triggering early stopping well before completion. The authors suggest that behavioral topology gets shaped more by the deployment harness than by the LLM underneath it. Paper: arxiv.org/abs/2608.23670 Chat with Paper: academy.dair.ai/papers/automat… 🔗 View Quoted Tweet

This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Automata, DAIR.AI, agent harness, finite-state machines.

Source attribution: curated from elvis (@omarsar0).