Reproduce “Attention-based Hierarchical Variational Information Bottleneck for Robust Multi-Agent Communication under Variable Bandwidth” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lukas Koch Vindbjerg, Qi Zhang, Yury Brodskiy, Lukas Esterle
AH-VIB models cooperative robot messages as ordered dimensions and trains agents to preserve estimated team value under truncated prefixes. In a simulated inspection task with 32-dimensional messages, it reports mean episode return of −0.2 ± 0.5 at bandwidth fraction ρ=0.10, versus −7.7 ± 24.3 for flat VIB. AH-VIB also leads at ρ=0.25, while CommNet leads at ρ=0.75 and 1.0. Results concern simulated prefix truncation, not real-world robustness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.