Metraipolis › archive › dc_82ce9cfd2268
RLCascadeRouter: Quality-Estimator-Free Cascade Routing via Reinforcement Learning — arXiv:2608.15817
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.15817
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T17:41:05.469Z
- SHA-256 of the archived text
3a82c097dcfe267b896c8c7e627863121b5dc0338e663de24e0a57e6be63c2ec- Stable URL of this page
https://metraipolis.com/documento/dc_82ce9cfd2268
How to cite
RLCascadeRouter: Quality-Estimator-Free Cascade Routing via Reinforcement Learning — arXiv:2608.15817 — arXiv (en), https://arxiv.org/abs/2608.15817 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_82ce9cfd2268 (Metraipolis archive dc_82ce9cfd2268, sha256 3a82c097dcfe267b896c8c7e627863121b5dc0338e663de24e0a57e6be63c2ec)
BibTeX
@misc{dc82ce9cfd2268,
title = {RLCascadeRouter: Quality-Estimator-Free Cascade Routing via Reinforcement Learning — arXiv:2608.15817},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.15817},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_82ce9cfd2268; sha256 3a82c097dcfe267b896c8c7e627863121b5dc0338e663de24e0a57e6be63c2ec},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15817. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
The growing ecosystem of large language models (LLMs) offers huge potential to optimize performance-cost trade-offs. However, their heterogeneous capabilities and inference costs make efficiently routing queries a significant challenge. Existing paradigms are inflexible: one-shot routers commit before observing responses, whereas conventional cascades stop adaptively but follow a fixed model order. Cascade routing removes both restrictions by reconsidering whether to stop or invoke another model after each response. Current methods use a predict-then-optimize pipeline estimating response quality and future model utility. However, prediction loss for quality or utility is not equivalent to routing-decision loss. A lower prediction error does not necessarily yield a better action; a small boundary-crossing error can reverse a ``stop'' or model-selection decision. Therefore, we propose RLCascadeRouter, a quality-estimator-free framework that formulates cascade routing as a Markov decision process with actions comprising ``stop'' and model selection. It uses trajectory returns and advantages to directly optimize the performance-cost objective. Its Cascade Policy Network models candidate complementarity for model selection and remaining-action value for stopping, eliminating independent post-hoc response-quality estimators. Evaluated across ten LLMRouterBench benchmarks with thirteen LLMs, RLCascadeRouter outperforms strong baselines and achieves superior performance-cost trade-offs. It incorporates unseen models without retraining, and ablation studies validate both policy components. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.15817 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T16:29:13.833Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).