Main Conference Day 1: Thursday 26th September
08.00 – 09.00: Registration and Morning Welcome Coffee
09.00 – 09.15: Legends in Quantitative Finance: Riccardo Rebonato
Introduction: Andrei Lyashenko
Riccardo Rebonato:
Professor – Climate Impact Institute Scientific and Programme Director, EDHEC Business School
Andrei Lyashenko:
Head of Market Risk and Pricing Models, Quantitative Risk Management (QRM), Inc.
09.15 – 10.00: Notation as a tool of thought? LLMs, cognitive biases, and logical fallacies.
In 1979, Kenneth E. Iverson, the inventor of APL (which became the progenitor of an entire family of programming languages, such as A+, k, and q) delivered his ACM Turing Award Lecture. He emphasised the “importance of nomenclature, notation, and language as tools of thought.” Iverson, along with other computer scientists, such as Bjarne Stroustrup and Guido van Rossum continued to work on programming languages – specialized notations that eschewed the limitations of natural (human) languages, thus facilitating complex and precise computation – and advancing science. In recent years, there has been a reversal in computer science: natural language became the tool of choice; the widely adopted large language models (LLMs) use natural language as input and output and work on natural language-based intermediate representations. Nonetheless, there is a certain unease about natural languages, which is probably a legacy of “Newspeak” described in George Orwell’s “Ninety-Eighty Four.” Not only can natural languages be used as a tool of manipulation (as in Orwell’s dystopian novel); they are inherently vulnerable to cognitive biases and logical fallacies, which may be an unavoidable feature of human existence. In this talk we explore the implications of this vulnerability. We also examine the significant role that cognitive biases and logical fallacies play in finance and economics. We compare and contrast LLMs and other directions in AI, such as reinforcement learning, where the emphasis is on provable optimality, which is missing in LLMs. Finally, we discuss various approaches to fixing this vulnerability.
Paul Bilokon:
CEO, Thalesians, Visiting Professor, Imperial College
10.00 – 10.45: Generative AI for Data Analysis
Ioana Boier:
11.15 – 12.00: Evaluating LLMs in Financial Tasks – Code Generation in Trading Strategies
In this paper, we perform a comprehensive evaluation of various Large Language Models (LLMs) for their efficacy in generating Python code specific to algorithmic trading strategies. Our study encompasses a broad spectrum of LLMs, including GPT-4-Turbo, Gemini-Pro, Mistral, Llama2, and Codellama, assessing their performance across a series of task-specific prompts designed to elicit precise code implementations for over various technical indicators commonly used in the financial trading sector.
A principal component of our methodology involves the creation of a detailed prompt structure that adapts to the unique capabilities of each LLM. For OpenAI’s Assistant AI, we leverage an intricate prompt design that integrates templated responses, zero-shot task- specific prompts, and prompt chaining to guide the models through a step-by-step reasoning process, ensuring the generation of executable and accurate Python code. This structured approach not only facilitates the model’s comprehension of the task at hand but also allows for the nuanced adaptation of prompts to cater to the distinct processing styles of different LLMs.
Our evaluation framework is grounded in a comparison against baseline results obtained from widely recognized libraries such as TALib, as well as a comprehensive Python implementation of the indicators. Through a meticulous process of parsing our code and constructing data frames that encapsulate function names, parameters, and documentation, we establish a foundational prompt that prompts LLMs to propose viable Python code implementations. This zero-shot task-specific approach is crucial in enabling the LLMs to methodically navigate through the tasks, thereby enhancing the accuracy and relevance of the generated code. The findings indicate that GPT-4-Turbo, Codellama-70B, and Gemini-Pro yield encouraging results relative to baseline computations, with GPT-4-Turbo achieving identical implementations to the baseline in certain instances.
Miquel Noguer Alonso:
Co – Founder and Chief Science Officer, Artificial Intelligence Finance Institute – AIFI
12.00 – 12.45: Integrating Large Language Models for Enhanced ESG and Climate Risk Analysis in Finance
In this session, we’re diving into how Large Language Models (LLMs) can be used in ESG and climate risk analysis for finance. We’ll break down how these AI tools help us sift through huge amounts of data to spot risk patterns and insights that might not be obvious at first glance. We’ll see firsthand the value of LLMs in understanding and managing the risks tied to ESG. Will look into how to leverage these technologies for a more informed decision-making approach in finance.
David Pacheco Aznar:
Founding Partner & Chief of AI Research and Development, Raven Risk AI
14.00 – 14.45: What-If Trade Entry from Email using LLMs
- Manually entering complex trades for what-if analysis is a slow and unreliable component of what is otherwise a highly effective process of trade execution
- We discuss the key techniques involved in building LLM-based trade entry from natural language trade description contained in unstructured email messages or threads
- We present specialized prompting and RAG techniques for dealing with complex trade definition formats and logic
- GPT-4o, LLAMA 2/3 and Mixtral models are compared on examples from all major asset classes
- Open-source code for all examples and tests will be provided
Alexander Sokol:
Executive Chairman and Head of Quant Research, CompatibL
14.45 – 15.30: How to apply LLM Agents in Finance?
Hanane Dupouy:
Director – Algorithmic Trader, Societe Generale Corporate and Investment Banking
16.00 – 16.45: LLM Adaptation to Financial Tasks
- We investigate how foundational LLMs can be adapted to specific financial tasks and what are the issues and challenges
- We use Parameter-Efficient Finetuning (PEFT) such as LoRA and QLoRA and extensive textual datasets from LSEG/Refinitiv to finetune LlaMa 3
- We consider two main business cases: analysis of analysts’ call transcripts and aggregate analysis of financial news.
- We compare the performance of finetuned and foundational LLMs in both contexts and provide answers to some of the challenges such as data pipeline, compute time and quality of the resulting finetuned LLMs
Svetlana Borovkova:
Head of Quantitative Modelling, Probability & Partners. Associate Prof, Vrije Universiteit Amsterdam
16.45 – 17.30: Pathwise Methods and Generative Models for Pricing and Trading
The deep hedging framework as well as related deep trading setups have opened new horizons for solving hedging problems under a large variety of models and market conditions. In this setting, generative models and pathwise methods rooted in rough paths have proven to be a powerful tool from several perspectives. At the same time, any model – a traditional stochastic model or a market generator – is at best an approximation of market reality, prone to model-misspecification and estimation errors. In a data-driven setting, especially if sample sizes are limited by constraints, the latter issue becomes even more prevalent, which we demonstrate in examples. This raises the question, how to furnish a modelling setup (for deriving a strategy) with tools that can address the risk of the discrepancy between model and market reality, ideally in a way that is automatically built in the setting. A combination of classical and new tools yields insights into this matter.
Blanka Horvath:
Associate Professor in Mathematical and Computational Finance, University of Oxford
09.00 – 09.15: Legends in Quantitative Finance: Riccardo Rebonato
Introduction: Andrei Lyashenko
Riccardo Rebonato:
Professor – Climate Impact Institute Scientific and Programme Director, EDHEC Business School
Andrei Lyashenko:
Head of Market Risk and Pricing Models, Quantitative Risk Management (QRM), Inc.
09.15 – 10.00: Volatility Shape-Shifters: arbitrage-free transformations of a volatility surface
- No Arbitrage & Convex order
- Optimal transport
- One-X & Ohlin’s theorem
- Convex Order generating/preserving transformations
- Time extrapolation:
- No market information
- Regeneration with Memory
- Conclusions
Valer Zetocha:
Senior Quantitative Analyst, ED, Julius Baer
10.00 – 10.45: From Local Volatility to Local Time
- Revisit of probabilistic techniques for local volatility models
- Generic formulation of the probability density function (pdf)
- Applications:
- Extension of Quadratic Volatility with additional control over higher order moments
- Regime change, Singular Local Volatility and connection to Local Time
- Closed form solutions for option pricing
Dominique Bang:
Managing Director, FICC Quantitative Modelling Lead, Bank Of America
11.15 – 12.00: Next Gen Finite Difference Grids
- Stability and accuracy of fd solution: myths, facts and orthodoxy
- Discrete consistency: backwards, forwards and Dupire fd solution.
- local volatility and absence of arbitrage
- Dividends.
- stochastic volatility
- jumps
- american options
- the spx/vix conundrum again again
Jesper Andreasen:
Head of Quantitative Analytics, Verition Fund Management LLC
12.00 – 12.45: P vs Q ~ drift vs vol
P vs Q can be described as modelling drift vs modelling volatility. Under Q-measure, drifts are fixed and the bulk of research focuses on modelling volatility and correlation. Under P-measure, the opposite is the case: it is (relatively) easy to estimate historical volatilities and correlations, but specification of the long-term drifts is a problem. One way of dealing with it is to introduce historical regime switch model, removing the necessity to provide a single drift estimate. We consider such a model for an asset-like product modelling.
Andrey Chirikhin:
Head of Structured Credit QA, Barclays Investment Bank
14.00 – 14.45: Transaction costs for cash-settled options, revisited
- Option exercise optimality is in the eye of the (be-)holder
- Spot not the value of the hedge
- Getting in is not for free
- Cash-settled options, à la Leland, but not quite.
Peter Jaeckel:
Independent financial mathematics and analytics consultant. OTC Analytics
14.45 – 15.30: Towards the lower bound for Bermudans
- An achievable lower bound for a Bermudan is not always the max European
- Finding a robust (model-free) lower bound for Bermudans is surprisingly difficult
- It is important for designing “Bermudan discount” adjustment models prevalent in the industry for the last 30+ years
- We make some progress towards this goal
Vladimir Piterbarg:
MD, Head of Quantitative Analytics and Quantitative Development, NatWest Markets
16.00 – 16.45: The Geometry of Finance
- Parametric models may prevent static arbitrage but their recalibration often produce dynamic arbitrage
- The geometry of hedging: dependency on the target instrument, the hedging instruments and the model
- The geometry of investing: alpha explained by the Numeraire Portfolio and consequences
Bruno Dupire:
Head of Quantitative Research, Bloomberg
16.45 – 17.30: Arbitrage-free P and Q dynamics of Nelson-Siegel factor HJM model with stochastic volatility
- Dynamics of Factor HJM model under P and Q measures
- Risk-premia in Nelson-Siegel HJM model
- Calibration to swaptions and options on rates futures under Q measure
- Forecast and long-term simulation of model dynamics under P and Q measures
Artur Sepp:
Head Quant, LGT Bank
Parviz Rakhmonov:
Vice President, Quantitative Analyst, Citibank
09.00 – 09.15: Legends in Quantitative Finance: Riccardo Rebonato
Introduction: Andrei Lyashenko
Riccardo Rebonato:
Professor – Climate Impact Institute Scientific and Programme Director, EDHEC Business School
Andrei Lyashenko:
Head of Market Risk and Pricing Models, Quantitative Risk Management (QRM), Inc.
09.15 – 10.45: Extended Session: On the use of Autoencoders in Risk-Neutral Yield Curve Modeling
(based on joint work by Andrei Lyashenko, Fabio Mercurio and Alexander Sokol)
Abstract:
It is well known that yield curves have low effective dimensionality and can be accurately represented using very few latent variables. The recent extension to nonlinear representations by means of autoencoders provided further improvement in accuracy compared to the classical linear representations such as PCA. In this presentation we discuss whether and how an autoencoder trained to historical yield curve data can be used in risk-neutral modeling of a yield curve. We propose an approach for building pricing term structure models consistent with an empirically estimated yield curve autoencoder.
Bullets:
- Yield curve autoencoders and static no-arbitrage condition
- Statically arbitrage-free yield curve bootstrapping
- The case of affine (linear) yield curve autoencoders
- Constructing autoencoder-consistent term structure models
Andrei Lyashenko:
Head of Market Risk and Pricing Models, Quantitative Risk Management (QRM), Inc.
Alexander Sokol:
Executive Chairman and Head of Quant Research, CompatibL
11.15 – 12.00: What is the impact of physical climate risk on equity valuation?
- Transition risk vs physical risk: which one is more important?
- Plausible ranges of equity revaluation.
- The importance of discounting.
- Are damages already in the prices?
- Headwind versus Minsky moment.
Riccardo Rebonato:
Professor – Climate Impact Institute Scientific and Programme Director, EDHEC Business School
12.00 – 12.45: Guilt-free Cloud-based Production-grade Live Risk in Python
Dmitri Goloubentsev:
CTO, Head of Automatic Adjoint Differentiation, Matlogica
14.00 – 14.45: XVA Acceleration: the Nasdaq case study
- Strengths and pitfalls of Machine Learning for XVA
- A practical solution with Chebyshev Tensors: calibration and daily use
- The problem of dimensionality
- XVA sensitivities
- Case study: the Nasdaq XVA accelerator
Ignacio Ruiz:
14.45 – 15.30: Wrong-way risk and liquidity
Matthias Arnsdorf:
Global head of Counterparty Credit Risk Quantitative Research, J.P. Morgan
16.00 – 16.45: Credit Exposure to Leveraged Counterparties
We present a model for calculating credit exposure to leveraged collateralized counterparties. The model is based on the observation that a large margin call can cause default of such a counterparty. The proposed model of leveraged wrong-way risk (WWR) can be easily implemented over the existing exposure simulation framework. We also show how to incorporate conventional WWR using a local Gaussian approximation. Finally, we extend the expected exposure scaling approach proposed in Andersen et al. (2017) for calculating expected exposure in the presence of initial margin to the case of leveraged WWR.
Michael Pykhtin:
Manager, Quantitative Risk, U.S. Federal Reserve Board
16.45 – 17.30: Overcoming Markowitz’s Instability with the Help of the Hierarchical Risk Parity (HRP)
Download Event PDF