The results establish model substitution as a deployment-conditioned action rather than a property implied by a tier label, and they provide a practical sequence for large-scale workflow routing: calibrate the action, resolve its role-conditioned effect, and then optimize its placement.
Abstract
Large multi-call LLM systems pose a scientific problem that query-level routing does not capture: the value of a model depends on where it enters a dependent computation and on the deployment that surrounds that call. Existing routers typically decide \emph{where} to spend a stronger model while treating the benefit of the substitution itself as known. We separate these two decisions through a predicate-action factorization and evaluate it in controlled solve-merge-verify workflows spanning 8-64 solve calls and four three-tier model ladders. The resulting evidence reveals a consistent principle beneath apparently conflicting outcomes. On numeric frequency counting, all-strong reduces RMSE from 4.818 to 1.538 in the Mixed Qwen/GPT ladder, whereas the average Qwen-only ordering reverses. Input-matched interventions further show that the same medium-to-strong action has sharply different value across roles and scales. A semantic task-and-contract shift reverses the Mixed ordering again, while allocation ablations distinguish useful sparse placement from under-coverage and indiscriminate escalation. Together, these results establish model substitution as a deployment-conditioned action rather than a property implied by a tier label, and they provide a practical sequence for large-scale workflow routing: calibrate the action, resolve its role-conditioned effect, and then optimize its placement.
No single Large Language Model (LLM) is uniformly reliable across queries, motivating multi-model inference systems that either route among models or combine their outputs. However, routing stops after selecting an initial model, while dense collaboration invokes peers on every query. We show that collaboration is non-...
Norah Alballa, Wen-Xuan Zhang, Salma Kharrat et al.· 0 citations
Large language models (LLMs) increasingly construct multi-agent workflows that decompose a complex task and assign specialist agents from a pool. However, building such a workflow well remains challenging: how finely to divide the task, which agent to trust with each subtask, and when to create a new specialist are all...
Xue-Hang Guo, Hao-Yu Wang, Sheng-Yu Chen et al.· 0 citations
Experiments show that VERTICAL ROUTING outperforms the strongest baseline (RouterDC) by 2.0 points in the averaged metric, while significantly reducing token usage and large-model output share, and enhances stability by lowering the standard deviation by 87.5%.
Si Shen, Pei-Jun Shen, Dan-Hao Zhu· Transactions of the Associat...· 0 citations
RAILS is presented, a retrieval-augmented incremental LLM clusterer that turns clustering into a simple loop over a growing label pool and scales through document batching with bounded concurrency.
Armin Oliya, Aleksandra Sawczuk, Radosław Białobrzeski· arXiv.org· 0 citations
This work introduces the Differential Reasoning Router (DRR), a cost-aware framework for cold-start LLM annotation that jointly optimizes model selection and human escalation, enabling a gradual shift from human-heavy cold-start annotation toward high-confidence automated routing.
Cheng Lyu, Jingyu Zhang, Vinny DeGenova et al.· 0 citations