# Cascade-Aware Multi-Agent Routing: Spatio-Temporal Sidecars and Geometry-Switching

arXiv: https://arxiv.org/abs/2603.17112 · HTML: https://mlsanity.com/paper/2603.17112/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 37 references that are in the map.
- First posted on arXiv 17 Mar 2026.

## Builds on (most cited first, 20 of 20)

- [AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation](https://mlsanity.com/paper/2308.08155/) (2023)
- [Graph WaveNet for Deep Spatial-Temporal Graph Modeling](https://mlsanity.com/paper/1906.00121/) (2019)
- [Temporal Graph Networks for Deep Learning on Dynamic Graphs](https://mlsanity.com/paper/2006.10637/) (2020)
- [Inductive Representation Learning on Temporal Graphs](https://mlsanity.com/paper/2002.07962/) (2020)
- [EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs](https://mlsanity.com/paper/1902.10191/) (2019)
- [Hyperbolic Neural Networks](https://mlsanity.com/paper/1805.09112/) (2018)
- [Hyperbolic Graph Convolutional Neural Networks](https://mlsanity.com/paper/1910.12933/) (2019)
- [Learning Multiagent Communication with Backpropagation](https://mlsanity.com/paper/1605.07736/) (2016)
- [Representation Learning for Dynamic Graphs: A Survey](https://mlsanity.com/paper/1905.11485/) (2019)
- [TarMAC: Targeted Multi-Agent Communication](https://mlsanity.com/paper/1810.11187/) (2018)
- [Learning Attentional Communication for Multi-Agent Cooperation](https://mlsanity.com/paper/1805.07733/) (2018)
- [Towards a Science of Scaling Agent Systems](https://mlsanity.com/paper/2512.08296/) (2025)
- [Constant Curvature Graph Convolutional Networks](https://mlsanity.com/paper/1911.05076/) (2019)
- [Mixed-curvature Variational Autoencoders](https://mlsanity.com/paper/1911.08411/) (2019)
- [Discrete-time Temporal Network Embedding via Implicit Hierarchical Learning in Hyperbolic Space](https://mlsanity.com/paper/2107.03767/) (2021)
- [Hyperbolic Variational Graph Neural Network for Modeling Dynamic Graphs](https://mlsanity.com/paper/2104.02228/) (2021)
- [A Review of Self-Exciting Spatio-Temporal Point Processes and Their Applications](https://mlsanity.com/paper/1708.02647/) (2017)
- [Learning to Route LLMs from Bandit Feedback: One Policy, Many Trade-offs](https://mlsanity.com/paper/2510.07429/) (2025)
- [OpenCog Hyperon: A Framework for AGI at the Human Level and Beyond](https://mlsanity.com/paper/2310.18318/) (2023)
- [Instantiation](https://mlsanity.com/paper/1808.01662/) (2018)

## Built on by (most cited first, 1 of 1)

- [Learning When to Act: Interval-Aware Reinforcement Learning with Predictive Temporal Structure](https://mlsanity.com/paper/2603.22384/) (2026)

## Abstract (arXiv)

Advanced AI reasoning systems route tasks through dynamic execution graphs of specialized agents. We identify a structural blind spot in this architecture: schedulers optimize load and fitness but lack a model of how failure propagates differently in tree-like versus cyclic graphs. In tree-like regimes, a single failure cascades exponentially; in dense cyclic regimes, it self-limits. A geometry-blind scheduler cannot distinguish these cases. We formalize this observability gap as an online geometry-control problem. We prove a cascade-sensitivity condition: failure spread is supercritical when per-edge propagation probability exceeds the inverse of the graph's branching factor (p > e-γ, where γ is the BFS shell-growth exponent). We close this gap with a spatio-temporal sidecar that predicts which routing geometry fits the current topology. The sidecar comprises (i) a Euclidean propagation scorer for dense, cyclic subgraphs, (ii) a hyperbolic scorer capturing exponential risk in tree-like subgraphs, and (iii) a compact learned gate (133 parameters) that blends the two scores using topology and geometry-aware features. On 250 benchmark scenarios spanning five topology regimes, the sidecar lifts the native scheduler's win rate from 50.4% to 87.2% (+36.8 pp). In tree-like regimes, gains reach +48 to +68 pp. The learned gate achieves held-out AUC = 0.9247, confirming geometry preference is recoverable from live signals. Cross-architecture validation on Barabasi-Albert, Watts-Strogatz, and Erdos-Renyi graphs confirms propagation modeling generalizes across graph families.