Pareto-Front Agentic RL with Dynamic Preference Conditioning for Cost–Risk–Success Trade-offs in Web Tasks
Frontiers in Applied Physics and Mathematics · 2026-03-15 · 3 (2) · 25-33
DOI: 10.71465/fapm749 ↗At a glance
Journal article in Frontiers in Applied Physics and Mathematics, published online on 2026-03-15. Written by Marco Rossi and 2 co-authors. The full text is freely available (open access, CC BY 4.0).
Abstract (original language)
Web agent deployment requires navigating trade-offs among success rate, monetary/API costs, latency, and failure risk, which vary by user and scenario. We propose a preference-conditioned multi-objective RL framestudy that learns a Pareto set of policies for web tasks. The method trains a single agent conditioned on a preference vector w over (success, cost1_11, cost2_22, …, risk), using (i) multi-gradient conflict resolution to stabilize updates across objectives, and (ii) Pareto replay that balances samples from distinct regions of the frontier. To ensure tail-risk control, the risk objective is defined as quantile-based failure loss (e.g., 90th/95th percentile). We recommend benchmarking on 1,200+ tasks across 40–70 site templates, sweeping preference vectors to obtain a Pareto curve and reporting hypervolume improvement, frontier coverage, and policy switching stability. This approach enables “one model, many operating points,” supporting practical deployment where budgets and risk tolerance change dynamically.
Cite
APA
Rossi, M., Bianchi, G., & Conti, A. (2026). Pareto-Front Agentic RL with Dynamic Preference Conditioning for Cost–Risk–Success Trade-offs in Web Tasks. Frontiers in Applied Physics and Mathematics, 3(2), 25-33. https://doi.org/10.71465/fapm749
Vancouver
Rossi M, Bianchi G, Conti A. Pareto-Front Agentic RL with Dynamic Preference Conditioning for Cost–Risk–Success Trade-offs in Web Tasks. Frontiers in Applied Physics and Mathematics. 2026;3(2):25-33. doi:10.71465/fapm749
BibTeX
@article{Rossi2026,
author = {Rossi, Marco and Bianchi, Giulia and Conti, Alessandro},
title = {Pareto-Front Agentic RL with Dynamic Preference Conditioning for Cost–Risk–Success Trade-offs in Web Tasks},
journal = {Frontiers in Applied Physics and Mathematics},
year = {2026},
volume = {3},
number = {2},
pages = {25--33},
doi = {10.71465/fapm749}
}
EUVIMED