AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

MixRoute: Rethinking Single-Distribution Training for Generalizable Neural Routing

Hang Yi, Ziwei Huang, Yining Ma, Zhiguang Cao

NeurIPS 2026 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Zero-shot robustness and transfer · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Challenges the assumption that single-distribution training gives weak cross-distribution generalization in neural combinatorial optimization for routing problems, using the MixRoute approach.

Problem

Generalization of neural combinatorial optimization (NCO) routing solvers across diverse data distributions; existing fixes rely on complex multi-distribution training or meta-learning.

Approach

Questions the presumption that single-distribution training yields a weak zero-shot baseline, suggesting that a simpler single-distribution training scheme can generalize; the specific mechanism is only partially described in the abstract provided.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace POMO's normalization with MixNorm (IN+LN+BN), trained on uniform onlyPOMO · TSP100 zero-shot avgoptimality gap: 1.70% -> 0.87%Improved
replace POMO's normalization with MixNormPOMO · CVRP100 zero-shot avgoptimality gap: 2.83% -> 0.72%Improved
replace POMO's normalization with MixNormPOMO · CVRPTW100 zero-shot avgoptimality gap: 4.37% -> 2.65%Improved
replace POMO's normalization with MixNormPOMO · CVRPTW100 time-window Width shiftoptimality gap: 7.73% -> 12.55%Hurt
apply MixNorm to MatNetMatNet · ACVRP100 demand shiftsoptimality gap: -1.06% -> -1.63%Improved
use MixNorm on CVRPLIB real-world setsPOMO · CVRPLIB seven setsoptimality gap: 11.80% -> 7.94%Improved
use NoNorm instead of BN as third constituent of MixNorm (ablation)MixRoute · CVRP100optimality gap: 0.88% -> 0.95%No clear effect
use RMSNorm instead of BN as third constituent (ablation)MixRoute · CVRP100optimality gap: 0.88% -> 1.41%Hurt
use output-level MixNorm instead of Switchable Norm statistic mixing (ablation)MixRoute · CVRP100optimality gap: 1.68% -> 0.88%Improved
use MixRoute instead of multi-distribution AMDKD-POMOMixRoute · TSP100 zero-shot avgoptimality gap: 1.05% -> 0.87%Improved

Surprising findings

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