AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Budget-Efficient Attacks and Robustness Training for Cooperative MARL

junyong jiang, Xin Yuan, Longhe Lin, Songze Li, Lu Dong

ICML 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Attacks and defenses in sequential settings

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 BHEA-AT instead of ROMANCE or WALL for adversarial trainingQMIX victims · SMAC MMM under BHEA attacktest win rate (%): 80.2 / 63.1 -> 99.2Improved
adversarial training with BHEA-AT instead of vanilla trainingQMIX · SMAC MMM under EGA attacktest win rate (%): 53.1 -> 98.1Improved
adversarial training with BHEA-ATQMIX · SMAC natural execution (no attack)test win rate (%): 99.0 -> 99.4 on MMMNo clear effect
BHEA-AT instead of WALL under EGA attack on 8mQMIX · SMAC 8m under EGA attacktest win rate (%): 77.1 -> 78.3No clear effect
BHEA-AT instead of ROMANCE under EGA attack on 2s3zQMIX · SMAC 2s3z under EGA attacktest win rate (%): 94.2 -> 95.3No clear effect
use BHEA-AT instead of ROMANCE for training costadversarial training · SMAC MMMtraining time (hours): 30h -> 18hImproved
use BHEA-AT instead of WALL for training costadversarial training · SMAC MMMtraining time (hours): 47h -> 18hImproved
replace learned victim selector with uniform random selection (-Selector) (ablation)BHEA-AT · SMAC 1c3s5z and MMMwin rate under attackHurt
replace learned action attacker with random actions (-Attacker) (ablation)BHEA-AT · SMAC 1c3s5z and MMMwin rate under attackHurt
replace learned action attacker with lowest-Q action (-Attacker + WorstQ) (ablation)BHEA-AT · SMAC 1c3s5z and MMMwin rate under attackMixed
use full BHEA instead of ablated selector or attacker variants (ablation)BHEA attack · SMAC 1c3s5z and MMMattack strength (victim win rate)Improved
increase per-timestep attack capacity K (ablation)BHEA · SMAC 1c3s5z (B=4), MMM (B=3)victim win rateImproved

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