AI Paper Atlas

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

Triple-BERT: Do We Really Need MARL for Order Dispatch on Ride-Sharing Platforms?

Zijian Zhao, Sen Li

ICLR 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: MARL benchmarks and grouping · 3 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
add MARL (IDDQN) stage-1 pre-training of feature extractors before centralized TD3 (ablation)Triple-BERT · Manhattan multi-periodaverage reward: 10665.02 -> 14730.48Improved
add positive normalization to QK-attention (ablation)Triple-BERT · Manhattan multi-periodaverage reward: 10839.33 -> 14730.48Improved
use centralized single-agent TD3 with BERT instead of best MARL baseline (CEVD)Triple-BERT · Manhattan multi-periodaverage reward: 13157.96 -> 14730.48Improved
use centralized Triple-BERT instead of CEVDTriple-BERT · Manhattan FHV generalizationdelivery time: 12.33 -> 13.07Hurt
use Triple-BERT instead of BMG-QTriple-BERT · Manhattan multi-perioddelivery/detour time: 10.57 -> 11.53; 1.90 -> 2.52Hurt
use Triple-BERT instead of CEVDTriple-BERT · Manhattan FHVservice rate: 0.80 -> 0.88Improved

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