Skip to content

GeomHerd: A Forward-looking Herding Quantification via Ricci Flow Geometry on Agent Interactive Simulations

Home / Papers / GeomHerd: A Forward-looking Herding Quantification via Ricci Flow Geometry on Agent Interactive Simulations

GeomHerd: A Forward-looking Herding Quantification via Ricci Flow Geometry on Agent Interactive Simulations

Authors: Yang Lake, Junwei Su, Jingfeng Zeng, Wenhao Lu, Xingzhi Qian, Weitong Zhang, Chuan Wu, Dunhong Jin Date: 2026-05-12 Paper ID: openalex:2605.11645

Summary

GeomHerd is a forward-looking framework designed to detect herding behavior in multi-agent financial simulations by analyzing the geometric properties of agent-interaction graphs. Instead of relying on lagging price-correlation statistics, the framework uses discrete Ollivier-Ricci curvature to identify structural topological changes indicating emergent coordination. Experiments on a continuous-spin agent-based substrate demonstrate that the method anticipates herding onset significantly earlier than traditional baselines, with transferability to other agent-driven models.

Key Contributions

  • Introduces GeomHerd, a geometric framework that quantifies herding behavior by calculating discrete Ollivier-Ricci curvature on agent-interaction graphs, enabling forward-looking detection of systemic risk.
  • Establishes a mean-field theoretical bridge between the proposed graph-theoretic curvature metric and the classical CSAD (Cross-Sectional Absolute Deviation) herding statistic.
  • Demonstrates that GeomHerd identifies herding signals significantly earlier than price-correlation baselines, showing a 272-step lead on the continuous-spin agent-based substrate and 40-step lead on co-firing trajectories.

Open Questions & Future Work

Key Concepts

  • geomherd: A geometric framework for quantifying herding and systemic risk by measuring discrete Ollivier-Ricci curvature on agent-interaction graphs.

Archivist Review

I approved the GeomHerd framework as a novel geometric approach to systemic risk signaling in multi-agent systems, and the open question regarding its theoretical grounding in macroscopic indicators. These represent distinct, reusable contributions to time-series analysis and collective dynamics modeling. No datasets were approved as none were presented as novel benchmarks or research assets beyond standard simulation substrates.

Approved Concepts

  • GeomHerd: Introduces a novel geometric approach to quantifying emergent systemic behavior by measuring Ricci curvature on agent-interaction graphs rather than lagging statistics.

Approved Open Questions

  • Rigorous Proof for Ricci-CSAD Bridge: This theoretical link is essential for anchoring new geometric indicators to established, widely-used financial stability measures, providing formal validity to the interpretation of these new signals within the existing finance literature.

Metadata & Links

url
https://arxiv.org/abs/2605.11645
paper_id
2605.11645
paper_source
openalex
domain
time-series
tags
time-series-forecastingmulti-agent-systemsfinancial-forecastinggraph-neural-networks
architectures
datasets
concept_slugs
geomherd
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-05-15T06:15:52Z