AI Paper Atlas

Research map / Multi-agent systems and games / Cooperative multi-agent RL

Multi-Action Self-Improvement For Neural Combinatorial Optimization

Laurin Luttmann, Lin Xie

ICLR 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: MARL benchmarks and grouping · 2 citations (OpenAlex)

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
use multi-agent set-based policy (MACSIM) instead of single-action SLIM, samplingSLIM · FJSP 10x5makespan: 98.74 -> 97.64Improved
use MACSIM instead of SLIM, samplingSLIM · FFSP 50x4x3makespan: 48.13 -> 47.74Improved
use MACSIM instead of SLIM, samplingSLIM · HCVRP 100x3max route length: 10.81 -> 10.67Improved
MACSIM versus OR-Tools on FJSP 20x5 with samplingOR-Tools · FJSP 20x5makespan: 188.15 -> 188.10Improved
MACSIM versus OR-Tools on FJSP 15x10 with samplingOR-Tools · FJSP 15x10makespan: 143.53 vs 145.95Hurt
add autoregressive agent sampling instead of fixed order (ablation)MACSIM · FJSP 15x10makespan: 155.13 -> 145.95Improved
add autoregressive agent sampling instead of random order (ablation)MACSIM · FJSP 15x10makespan: 150.72 -> 145.95Improved
add skip token (ablation)MACSIM · FJSP 15x10makespan: 159.65 -> 145.95Improved
add skip token (ablation)MACSIM · FJSP 15x10inference time (s): 3.89 -> 6.19Hurt
add skip token (ablation)MACSIM · FJSP 10x5makespan: 98.69 -> 97.64Improved
use set-based cross-entropy loss instead of single-agent cross-entropy (ablation)MACSIM · FFSP 50x12makespan: 49.58 -> 47.74Improved
use set-based cross-entropy instead of L_PL (ablation)MACSIM · FJSP 10x5makespan: 97.99 -> 97.64Improved

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