MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641115864 A) filed by Vellore Institute Of Technology on September 28, 2026, for Semantic-To-Causal State Reconstruction With Dependency-Aware Control Of Agent Execution.
Inventors include Kalaiselvan N; and K. Sabari Ganesan.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: ABSTRACT The present invention provides a dual-engine control architecture for causally consistent neural-state recovery and dependency-aware autonomous agent execution. A Semantic-to-Causal Boundary Engine identifies a semantic invalidity in an autoregressively generated output, constructs a semantic dependency closure, maps the closure to token positions, and determines quantitative causal influence of prior hidden activations using counterfactual interventions. An earliest causally safe recovery boundary is derived from the causal influence and used to selectively truncate key-value (KV) cache tensors while preserving unaffected neural context without full prompt-prefix recomputation. A Dependency-Aware Transactional Agent Execution Gate intercepts generated agent actions in a non-committed state, extracts external parameter dependencies, and selectively evaluates environmental state drift across dependency-relevant resources. External invalidations are mapped to corresponding neural token positions to trigger targeted KV-cache recovery and parameter regeneration. Action execution is permitted using jointly validated external dependency consistency and internal neural causal safety, with version-bound cryptographic authorization tokens preventing execution following intervening state changes and thereby mitigating TOCTOU race conditions.
Disclaimer: Curated by HT Syndication.
