Representations
What does company information reveal? Language models turn news and disclosures into comparable representations of business activities and risks.
Research
Measuring the economic structure behind financial risk
Company information describes what businesses do. Financial markets record how securities behave. This research connects the two, using language models and statistical methods to measure economic peers, shared risk and strategic competition.
Relationships
Which firms are economically connected? Optimal transport and statistical models identify information-based peers and shared financial exposure.
Decisions
What can be learned when return histories are incomplete? Models connect economic differences to portfolio risk, with explicit assumptions and limitations.
Applied artificial intelligenceCausal inferenceInformation geometryQuantitative financeReproducible research
2026Preprint; accepted for World Finance & Banking Symposium, Riga (2026)
Information Games: Strategic Crowding and Firm Repositioning in Language-Model Space
Marcus Gawronsky · Chun-Sung Huang
The question: How does competition change firms' responses to opportunity in language-model space?
2026Submitted / under review
Wasserstein-Barycentric Interaction Fields for Spatial Factor Models: Evidence from Language-Model Representations
Marcus Gawronsky · Chun-Sung Huang
The question: Can distributional similarity define economically meaningful interaction fields?
Methods: Optimal transport and Wasserstein geometry · Covariance and factor structure · Language-model representations
2026Submitted / under review
Portfolio Risk Bounds without Cross-Asset Return Covariances: Distributional Fields from Language-Model Representations
Marcus Gawronsky · Chun-Sung Huang
The question: Can portfolio risk be bounded without estimating every cross-asset covariance directly?
Methods: Covariance and factor structure · Language-model representations · Simulation and uncertainty · Optimisation and decisions
2024Preprint; presented at World Finance Conference, Malta (2025)
Systematic Covariance Envelopes from Wasserstein Geometry: Evidence from Language-Model Representations
Marcus Gawronsky · Chun-Sung Huang
The question: How much systematic covariance can be recovered from the geometry of probability-valued representations?
Methods: Optimal transport and Wasserstein geometry · Covariance and factor structure · Language-model representations
Method index
Where each idea appears
The same ideas recur across papers, software, writing and figures. Follow one method to everywhere it appears in the work.
Method 01
Optimal transport and Wasserstein geometry
The question: What does the geometry of probability distributions say about the firms they describe?
Method 02
Covariance and factor structure
The question: How much dependence between assets can be recovered, or bounded, without estimating it directly?
- Papers
- Wasserstein-Barycentric Interaction Fields for Spatial Factor Models: Evidence from Language-Model Representations
- Portfolio Risk Bounds without Cross-Asset Return Covariances: Distributional Fields from Language-Model Representations
- Systematic Covariance Envelopes from Wasserstein Geometry: Evidence from Language-Model Representations
- Projects
- Pricing Perspective
- Figures
- Principal components
Method 03
Language-model representations
The question: Can a model's representation of a firm become a measurable object in quantitative finance?
- Papers
- Wasserstein-Barycentric Interaction Fields for Spatial Factor Models: Evidence from Language-Model Representations
- Portfolio Risk Bounds without Cross-Asset Return Covariances: Distributional Fields from Language-Model Representations
- Systematic Covariance Envelopes from Wasserstein Geometry: Evidence from Language-Model Representations
- Projects
- Pricing Perspective
Method 04
Event studies and network econometrics
The question: What does the market learn when a network of financial relationships becomes public?
- Projects
- Precarious Papers
Method 05
Simulation and uncertainty
The question: How far can an estimate be trusted, and how does randomness accumulate into a law?
- Writing
- Beyond Average Revenue
Method 06
Optimisation and decisions
The question: Which choice should be made when objectives conflict?
- Figures
- Pareto frontier
Method 07
Reproducible research systems
The question: Can every result be traced to the data, code and environment that produced it?