aiqbench analysis
This research introduces ECO, a novel framework for Neural Combinatorial Optimization that integrates batched Direct Preference Optimization with a Mamba state-space model backbone. By decoupling trajectory generation from gradient updates and utilizing mixed encoder-decoder architectures, the approach significantly improves memory efficiency and throughput on long sequences without sacrificing solution quality at inference time. The study demonstrates superior performance over existing neural baselines in solving Traveling Salesperson Problem (TSP) instances while maintaining scalable hardware utilization.
Rethinking Efficiency in Neural Combinatorial Optimization: Batched Preference Optimization with Mamba
arXiv:2602.20730v3 Announce Type: replace Abstract: We study efficiency as a first-class objective in Neural Combinatorial Optimization (NCO) and present ECO, an efficient learning framework that combines batched preference optimization with a Mamba backbone. Instead of tightly interleaving every policy update with on-policy rollouts, ECO decouples trajectory generation from gradient updates through two stages: supervised warm-up on pre-computed solutions and iterative Direct Preference Optimization (DPO) on batched candidate sets generated by the current policy. We pair this learning pipeline with a mixed Mamba encoder-decoder that reduces memory growth on long sequences and improves hardware utilization. A local-search-guided bootstrapping strategy is further used during training to widen preference margins and stabilize iterative improvement. Importantly, local search is only used to construct stronger preference pairs during training and is never invoked at inference time. On TSP and CVRP, ECO achieves the strongest overall performance among the compared neural baselines while also delivering clear advantages in memory usage and throughput. We provide additional analysis on memory scaling, throughput, and the contribution of each design component.
Abstract:We study efficiency as a first-class objective in Neural Combinatorial Optimization (NCO) and present ECO, an efficient learning framework that combines batched preference optimization with a Mamba backbone. Instead of tightly interleaving every policy update with on-policy rollouts, ECO decouples trajectory generation from gradient updates through two stages: supervised warm-up on pre-computed solutions and iterative Direct Preference Optimization (DPO) on batched candidate sets generated by the current policy. We pair this learning pipeline with a mixed Mamba encoder-decoder that reduces memory growth on long sequences and improves hardware utilization. A local-search-guided bootstrapping strategy is further used during training to widen preference margins and stabilize iterative improvement. Importantly, local search is only used to construct stronger preference pairs during training and is never invoked at inference time. On TSP and CVRP, ECO achieves the strongest overall performance among the compared neural baselines while also delivering clear advantages in memory usage and throughput. We provide additional analysis on memory scaling, throughput, and the contribution of each design component.
Machine Learning (cs.LG)
arXiv:2602.20730 [cs.LG]
(or arXiv:2602.20730v3 [cs.LG] for this version)
Full text under the publisher's open licence · 177 words
Related coverage
Intelligence is Free, Now What? <br> Data Systems for, of, and by Agents
BAIR Blog · Jul 7, 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026
WeCon: An Efficient Weight-Conditioned Neural Solver for Multi-Objective Combinatorial Optimization Problems
arXiv · Jul 28, 2026
AutoPref: Automatic Discovery of Task-Specific Preference Objectives for Neural Combinatorial Optimization
arXiv · Jul 31, 2026