analiza aiqbench
This research presents a significant architectural shift in industrial AI by consolidating complex business logic and SOPs directly into model parameters via Continual Pre-training, eliminating the latency and error-proneness of traditional modular pipelines. The approach demonstrates tangible performance gains in a global financial system, reducing end-to-end latency by over 50% while boosting intelligent resolution rates, effectively replacing brittle engineering workflows with unified cognitive reasoning.
How to Train a Real-World Silicon Concierge? Internalizing Complex Business Workflow to Only OneModel
arXiv:2608.20350v1 Announce Type: cross Abstract: Traditional industrial agents rely on modular pipelines, including Router, Retriever, Planner, Executor, Responder, Reviewer, and other components. These systems often fracture into a labyrinth of ad-hoc patches, leading to cascading errors and high latency. We propose OneModel, an applicable paradigm shift from external workflows to internalized knowledge representation. Unlike modular systems that slice fluid user intents into static steps, OneModel consolidates complex business logic and SOPs directly into the model parameters. Through Continual Pre-training (CPT) and logic-compilation SFT, we transform fragmented business rules into intuitive model reasoning within a unified attention space. Deployed in our global financial service system, OneModel effectively breaks the trade-off between latency, accuracy, and complexity. Online A/B testing demonstrates an end-to-end latency reduction of more than 50 percent, from 18.7 seconds to 8.0 seconds, while the Intelligent Resolution Rate (IRR) increases from 64.3 percent to 83.3 percent. The results show that OneModel can replace brittle engineering logic with internalized cognitive intuition, offering a scalable blueprint for transitioning industrial agents from complex, error-prone workflows to unified model architectures.
Abstract:Traditional industrial agents rely on modular pipelines, including Router, Retriever, Planner, Executor, Responder, Reviewer, and other components. These systems often fracture into a labyrinth of ad-hoc patches, leading to cascading errors and high latency. We propose OneModel, an applicable paradigm shift from external workflows to internalized knowledge representation. Unlike modular systems that slice fluid user intents into static steps, OneModel consolidates complex business logic and SOPs directly into the model parameters. Through Continual Pre-training (CPT) and logic-compilation SFT, we transform fragmented business rules into intuitive model reasoning within a unified attention space. Deployed in our global financial service system, OneModel effectively breaks the trade-off between latency, accuracy, and complexity. Online A/B testing demonstrates an end-to-end latency reduction of more than 50 percent, from 18.7 seconds to 8.0 seconds, while the Intelligent Resolution Rate (IRR) increases from 64.3 percent to 83.3 percent. The results show that OneModel can replace brittle engineering logic with internalized cognitive intuition, offering a scalable blueprint for transitioning industrial agents from complex, error-prone workflows to unified model architectures.
Accepted to the ACL 2026 Industry Track (Oral). To appear in Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Industry Track)
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2608.20350 [cs.CL]
(or arXiv:2608.20350v1 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 211 cuvinte
