A redundancy-based filtering Q-learning algorithm makes all agents converge almost surely to optimal values under Byzantine attacks by enforcing a new verifiable topological condition on the communication graph.
Minimal construction of graphs with maximum robustness
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.MA 2years
2026 2verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
SAC is a decentralized iterative filter-and-refine protocol that achieves (F+1)-robustness in LLM multi-agent systems, suppressing Byzantine influence and improving performance on reasoning benchmarks where prior methods fail.
citing papers explorer
-
Fully Byzantine-Resilient Distributed Multi-Agent Q-Learning
A redundancy-based filtering Q-learning algorithm makes all agents converge almost surely to optimal values under Byzantine attacks by enforcing a new verifiable topological condition on the communication graph.
-
Robust Multi-Agent LLMs under Byzantine Faults
SAC is a decentralized iterative filter-and-refine protocol that achieves (F+1)-robustness in LLM multi-agent systems, suppressing Byzantine influence and improving performance on reasoning benchmarks where prior methods fail.