AI Paper Atlas

Research map / Reinforcement learning / Offline RL

InfraRL: A Benchmark for Constrained Resource Allocation in Large-Scale Infrastructure Asset Management

Yantian Wang, Wenhao Li, Bo Jin

ICML 2026 · Reinforcement learning › Offline RL · approach: Diffusion models for offline decision making

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add heuristic imitation loss to CQLCQL · InfraRL matrix evalMatrix eval (normalized): 0.4302 -> 0.7082Improved
add heuristic guidance to CQLCQL · InfraRL matrix evalFQE score: -0.1043 -> -0.0911Improved
switch allocator from score greedy to knapsack (ablation)CQL · InfraRLImprove vs History: 0.430 -> 0.703Improved
switch allocator from score greedy to knapsack (ablation)CQL-Heuristic · InfraRLImprove vs History: 0.708 -> 1.000Improved
switch allocator from score greedy to knapsack (ablation)MultiTask-BC · InfraRLImprove vs History: -0.042 -> -0.021No clear effect
switch allocator from score greedy to knapsack (ablation)QMIX-CQL-MF · InfraRLImprove vs History: 0.085 -> 0.444Improved
use mean-field aggregation in QMIX-CQLQMIX-CQL · InfraRLImprove vs History: -0.067 -> 0.085Improved
use multi-agent QMIX-CQL instead of single-agent CQLoffline RL · InfraRLViolation rate: 0.007 -> 0.711Hurt
increase budget factor from 1x to 4x for heuristic-guided value learning (ablation)CQL-Heuristic · 100-year simulationfinal average health: 3.41 -> 3.78Improved
increase budget factor up to 4.0x for pure imitation (ablation)Discrete-BC · 100-year simulationhealthNo clear effect
use value learning (CQL, CPQ) instead of random policyoffline RL · 100-year simulationfinal average health: 2.60 -> 3.04Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy