ATLAS-RTC introduces token-level runtime control that detects and corrects LLM drift from structured output contracts during the forward pass.
March 31, 2026
Original Paper
ATLAS-RTC: Closing the Loop on LLM Agent Output with Token-Level Runtime Control
arXiv · 2603.27905
The Takeaway
It moves reliability from 'retry loops' (expensive/slow) to 'closed-loop control' (fast). By applying interventions like masking and rollback at each step, it improves tool-calling success rates by up to 37% and slashes latency in failure-heavy scenarios.
From the abstract
We present ATLAS-RTC, a runtime control system for autoregressive language models that enforces structured output during decoding. ATLAS-RTC monitors generation at each step, detects drift from output contracts using lightweight signals, and applies targeted interventions such as biasing, masking, and rollback. Unlike post-hoc validation or static constrained decoding, it operates in a closed loop, enabling correction before errors materialize. Across structured generation and tool-calling tasks