Saved in:
Bibliographic Details
Main Author: Taberner-Miller, Annette
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.00136
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911591054704640
author Taberner-Miller, Annette
author_facet Taberner-Miller, Annette
contents Multi-model LLM serving operates in a non-stationary, noisy environment: providers revise pricing, model quality can shift or regress without notice, and new models arrive regularly. More than a dozen recent methods have proposed learned routers to navigate the resulting quality--cost tradeoff across portfolios spanning a $\sim$530$\times$ cost range. Despite this activity, two gaps in the current solution space limit routing effectiveness under these conditions: no existing router enforces a dollar-denominated cost ceiling in closed loop over an open-ended request stream, and none provides principled online adaptation to post-deployment shifts in pricing or model quality. We present ParetoBandit, an open-source adaptive router built on cost-aware contextual bandits that addresses both gaps. Its core contributions are: (1) an online primal--dual budget pacer that enforces a per-request cost ceiling without a known horizon, and (2) geometric forgetting on sufficient statistics that gives the bandit bounded memory for tracking quality and cost shifts. A hot-swap model registry further supports runtime model changes with budget-controlled exploration. On 1,824 benchmark prompts with a three-model portfolio, the router maintains budget compliance within 0.4%, adapts to price and quality shifts with up to +0.071 quality lift, and integrates a cold-started model within $\sim$142 steps.
format Preprint
id arxiv_https___arxiv_org_abs_2604_00136
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ParetoBandit: Budget-Paced Adaptive Routing for Non-Stationary LLM Serving
Taberner-Miller, Annette
Machine Learning
Computation and Language
68T05, 62L05
I.2.6; I.2.11; C.4
Multi-model LLM serving operates in a non-stationary, noisy environment: providers revise pricing, model quality can shift or regress without notice, and new models arrive regularly. More than a dozen recent methods have proposed learned routers to navigate the resulting quality--cost tradeoff across portfolios spanning a $\sim$530$\times$ cost range. Despite this activity, two gaps in the current solution space limit routing effectiveness under these conditions: no existing router enforces a dollar-denominated cost ceiling in closed loop over an open-ended request stream, and none provides principled online adaptation to post-deployment shifts in pricing or model quality. We present ParetoBandit, an open-source adaptive router built on cost-aware contextual bandits that addresses both gaps. Its core contributions are: (1) an online primal--dual budget pacer that enforces a per-request cost ceiling without a known horizon, and (2) geometric forgetting on sufficient statistics that gives the bandit bounded memory for tracking quality and cost shifts. A hot-swap model registry further supports runtime model changes with budget-controlled exploration. On 1,824 benchmark prompts with a three-model portfolio, the router maintains budget compliance within 0.4%, adapts to price and quality shifts with up to +0.071 quality lift, and integrates a cold-started model within $\sim$142 steps.
title ParetoBandit: Budget-Paced Adaptive Routing for Non-Stationary LLM Serving
topic Machine Learning
Computation and Language
68T05, 62L05
I.2.6; I.2.11; C.4
url https://arxiv.org/abs/2604.00136