Interference-Aware K-Step Reachable Communication in Multi-Agent Reinforcement Learning

Ziyu Cheng, Jinsheng Ren [email protected] Zhouxian Jiang [email protected] Chenzhihang Li [email protected], Rongye Shi [email protected] Bin Liang [email protected] Jun Yang [email protected], [0.4em] School of Software, Beihang University · 2026

Read Interference-Aware K-Step Reachable Communication in Multi-Agent Reinforcement Learning by Ziyu Cheng, Jinsheng Ren [email protected] Zhouxian Jiang [email protected] Chenzhihang Li [email protected], Rongye Shi [email protected] Bin Liang [email protected] Jun Yang [email protected], [0.4em] School of Software, Beihang University (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15054.txt.

Description

Cheng and co-authors present IA-KRC, a communication scheme for multi-agent reinforcement learning. Agents must pick helpful partners under tight bandwidth and shifting network layouts. A K-step reachability rule limits messages to neighbors the agent can actually reach. A second module predicts interference and steers partner choice toward high utility with low clash. Tests in complex, fast-changing multi-agent settings show stronger and more lasting cooperation than prior methods. The authors also report better robustness and scaling.

Catalog record

CREATORZiyu Cheng, Jinsheng Ren [email protected] Zhouxian Jiang [email protected] Chenzhihang Li [email protected], Rongye Shi [email protected] Bin Liang [email protected] Jun Yang [email protected], [0.4em] School of Software, Beihang University
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE65836
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.15054

Browse the full library · About the Tower of Babel