Chern–Simons Reservoir Memory: Autonomous Connectivity Dynamics and Density Transport
The article explores how a reservoir preserves the history of input signals both in its density distribution and in its geometric degrees of freedom. A comparison of four mechanisms shows that current creates a separate connected component that alters the subsequent density response, while accounting for all measured features yields a small improvement in decoding pulse order.

Density and connection as two memory channels
The reservoir encodes input history jointly in density and geometric degrees of freedom. A conserved current links the continuous density on the internal manifold to an independently evolving Chern–Simons (CS) connection.

- The model considers continuous density on the internal manifold.
- Geometric degrees of freedom contribute to history encoding.
- A conserved current links the density and the CS connection.
- The CS connection evolves independently.
Reservoir memory includes the joint state of the density and the connection. A single scalar response does not capture this channel.
Four models on one manifold
The study compares four models on the same internal manifold. The control models separate the contributions of transport and independent connection evolution.
| Model | Role in the comparison |
|---|---|
| Reciprocal transport | Transport model on the same manifold |
| Instantaneous transverse reconstruction | Control with instantaneous reconstruction |
| Local nonlinear feedback | Control with local feedback |
| Coupled CS evolution | Independent connection evolution |
The comparison is meaningful with the same internal manifold and explicit control models. Otherwise, the contribution of the connection cannot be isolated.
Resetting the connection component
The current generates an exact connection component. Instantaneous transverse reconstruction and scalar local closure do not reproduce this component. Resetting the component while keeping the density and Gauss-law curl unchanged alters the future density response. The effect holds under the chosen gauge and drift law. A small effect persists with numerical refinement.
The memory test criterion includes resetting the connection component while holding the density and Gauss-law curl fixed. Persistence of the effect with numerical refinement supports this criterion.
Where the models agree and where they diverge
Scalar memory metrics and density traces for pulse order are similar across the four models. Circulatory and longitudinal traces differ.
| Feature | Result |
|---|---|
| Scalar memory metrics | Similar across the four models |
| Density traces for pulse order | Similar across the four models |
| Circulatory traces | Differ |
| Longitudinal traces | Differ |
| All measured features, pulse-order decoding | CS model: +6.5 ± 1.9 percentage points over the instantaneous transverse reconstruction control; comparable to local feedback; paired comparisons by seed |
The evaluation criterion includes circulatory and longitudinal traces and decoding across all measured features. Scalar memory metrics alone do not distinguish the models.
Latency and dependence on preparation
In the first cycle, the coupled CS model has lower latency than the instantaneous reconstruction and local feedback models. In subsequent cycles, its latency is higher. This indicates dependence on preparation.
The comparison criterion includes latency across cycles and preparation. The first cycle does not give the full picture.
Frequently asked questions
Related materials
All materials
STAR: Adaptive Normalization of Temporal Context and Service Graphs for End-to-End Incident Management
The STAR framework trains representations from metrics, logs, and traces, using learnable normalization that accounts for multiscale temporal dynamics and dependency structures between services. On two real-world datasets, this pipeline delivers consistent improvements in anomaly detection, failure ranking, and root cause analysis over state-of-the-art baselines.
10 November 2026
Validating Candidates Costs More Than Generating Them: Honeypots Against LLM Vulnerability Hunting
The study examines the cost asymmetry in agentic vulnerability discovery: generating candidate versions is cheap, while confirming them through reachability analysis, execution, and PoC development consumes most of the budget, making verification a potential standalone target for defense. The proposed RedHerring—repository-adapted decoys with a hidden certificate—reduced the number of real vulnerabilities found by 38.7–60.4% at equal cost across 33 OSS-Fuzz projects; even after the agent was warned, the reduction remained at 37.2%.
9 November 2026
Checking LLM Hallucinations via SQL-Based Source Representation Without Fine-Tuning
The study examines whether an LLM can apply a basic symbolic skill—building a SQL database from source documents—to ground its outputs in sources without supervision when detecting hallucinations. On RAGTruth and DiaHalu, this neuro-symbolic pipeline proves more accurate than direct prediction and comparable to state-of-the-art methods, without requiring domain-specific fine-tuning.
9 November 2026