Markus Leippold

• Professor
• Chair at University of Zurich

Long-form legal reasoning remains a key challenge for large language models (LLMs) in spite of recent advances in test-time scaling. We introduce LEXam, a novel benchmark derived from 340 law exams spanning 116 law school courses across a range of subjects and degree levels. The dataset comprises 4,886 law exam questions in English and German, incl...

Tracking financial investments in climate adaptation is a complex and expertise-intensive task, particularly for Early Warning Systems (EWS), which lack standardized financial reporting across multilateral development banks (MDBs) and funds. To address this challenge, we introduce an LLM-based agentic AI system that integrates contextual retrieval,...

Accurate identification of true versus false climate information in the digital age is critical. Misinformation can significantly affect public understanding and policymaking. Automated fact-checking seeks to validate claims against trustworthy factual data. This study tackles the challenge of fact-checking climate claims by leveraging the currentl...

Company transition plans toward a low-carbon economy are key for effective capital allocation and risk management. This paper proposes a set of 64 indicators to comprehensively assess transition plans and develops a Large Language Model-based tool to automate the assessment of company disclosures. We evaluate our tool with experts from 26 instituti...

Instruction Fine-tuning (IFT) can enhance the helpfulness of Large Language Models (LLMs), but it may lower their truthfulness. This trade-off arises because IFT steers LLMs to generate responses with long-tail knowledge that is not well covered during pre-training, leading to more informative but less truthful answers when generalizing to unseen t...

With new cycles of global environmental assessments (GEAs) recently starting, including GEO-7 and IPCC AR7, there is increasing need for artificial intelligence (AI) to support in synthesising the rapidly growing body of evidence for authors and users of these assessments. In this article, we explore recent advances in AI and connect them to the di...

Retrieval Augmented Generation (RAG) is widely employed to ground responses to queries on domain-specific documents. But do RAG implementations leave out important information or excessively include irrelevant information? To allay these concerns, it is necessary to annotate domain-specific benchmarks to evaluate information retrieval (IR) performa...

To handle the vast amounts of qualitative data produced in corporate climate communication, stakeholders increasingly rely on Retrieval Augmented Generation (RAG) systems. However, a significant gap remains in evaluating domain-specific information retrieval - the basis for answer generation. To address this challenge, this work simulates the typic...

The aim of the study was to find out how new technologies can reduce the negative effects of climate change in Switzerland and make our society and economy more resilient. The experts have high hopes for the latest developments in artificial intelligence (AI) and satellite-based earth observations to overcome the various climate challenges. In comb...

What are Africa's climate information needs? How do they relate to global con-cerns? How can AI help? This paper begins exploring these topics, with a specific focus on Ghana and South Africa. Our study involves climate-related questions from Google Trends and identifies macro clusters related to the broad themes of understanding climate change, it...

Large Language Models have made remarkable progress in question-answering tasks, but challenges like hallucination and outdated information persist. These issues are especially critical in domains like climate change, where timely access to reliable information is vital. One solution is granting these models access to external, scientifically accur...

Understanding how climate change affects us and learning about available solutions are key steps toward empowering individuals and communities to mitigate and adapt to it. As Large Language Models (LLMs) rise in popularity, it is necessary to assess their capability in this domain. In this study, we present a comprehensive evaluation framework, gro...

In the face of climate change, are companies really taking substantial steps toward more sustainable operations? A comprehensive answer lies in the dense, information-rich landscape of corporate sustainability reports. However, the sheer volume and complexity of these reports make human analysis very costly. Therefore, only a few entities worldwide...

This paper introduces a novel approach to enhance Large Language Models (LLMs) with expert knowledge to automate the analysis of corporate sustainability reports by benchmarking them against the Task Force for Climate-Related Financial Disclosures (TCFD) recommendations. Corporate sustainability reports are crucial in assessing organizations' envir...

We explore the performance of mixed‐frequency predictive regressions for stock returns from the perspective of a Bayesian investor. We develop a constrained parameter learning approach for sequential estimation allowing for belief revisions. Empirically, we find that mixed‐frequency models improve predictability, not only because of the combination...

Multi-task learning (MTL) aims at achieving a better model by leveraging data and knowledge from multiple tasks. However, MTL does not always work -- sometimes negative transfer occurs between tasks, especially when aggregating loosely related skills, leaving it an open question when MTL works. Previous studies show that MTL performance can be impr...

Large Language Models (LLMs) have made significant progress in recent years, achieving remarkable results in question-answering tasks (QA). However, they still face two major challenges at inference time: hallucination and outdated information. These challenges take center stage in critical domains like climate change, where obtaining accurate and...

Large Language Models (LLMs) have made significant progress in recent years, achieving remarkable results in question-answering tasks (QA). However, they still face two major challenges: hallucination and outdated information after the training phase. These challenges take center stage in critical domains like climate change, where obtaining accura...

Large language models (LLMs) have significantly transformed the landscape of artificial intelligence by demonstrating their ability in generating human-like text across diverse topics. However, despite their impressive capabilities, LLMs lack recent information and often employ imprecise language, which can be detrimental in domains where accuracy...

In this paper, we introduce an expert-annotated dataset for detecting real-world environmental claims made by listed companies. We train and release baseline models for detecting environmental claims using this new dataset. We further preview potential applications of our dataset: We use our fine-tuned model to detect environmental claims made in a...

We use BERT, an AI-based algorithm for language understanding, to quantify regulatory climate risk disclosures and analyze their impact on the term structure in the credit default swap (CDS) market. Risk disclosures can either increase or decrease CDS spreads, depending on whether the disclosure reveals new risks or reduces uncertainty. Training BE...

The climate impact of AI, and NLP research in particular, has become a serious issue given the enormous amount of energy that is increasingly being used for training and running computational models. Consequently, increasing focus is placed on efficient NLP. However, this important initiative lacks simple guidelines that would allow for systematic...

We propose a new method, variable subsample aggregation (VASA), for equity return prediction using a large‐dimensional set of factors. To demonstrate the effectiveness, robustness, and dimension reduction power of VASA, we perform a comparative analysis between state‐of‐the‐art machine learning algorithms. As a performance measure, we explore not o...

Disclosure of climate-related financial risks greatly helps investors assess companies’ preparedness for climate change. Voluntary disclosures such as those based on the recommendations of the Task Force for Climate-related Financial Disclosures (TCFD) are being hailed as an effective measure for better climate risk management. We ask whether this...

In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in sample estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty: non-standard error...

In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in sample estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty: non-standard error...

Investors and regulators require reliable estimates of physical climate risks for decision-making. While assessing these risks is challenging, several commercial data providers and academics have started to develop firm-level physical risk scores. We compare six physical risk scores. We find a substantial divergence between these scores, also among...

We add to the emerging literature on empirical asset pricing in the Chinese stock market by building and analyzing a comprehensive set of return prediction factors using various machine learning algorithms. Contrasting previous studies for the US market, liquidity emerges as the most important predictor, leading us to closely examine the impact of...

Our goal is to introduce CLIMATE-FEVER, a new publicly available dataset for verification of climate change-related claims. By providing a dataset for the research community, we aim to help and encourage work on improving algorithms for retrieving climate-specific information and detecting fake news in social and mass media to reduce the impact of...

Climate change communication in the mass media and other textual sources may affect and shape public perception. Extracting climate change information from these sources is an important task, e.g., for filtering content and e-discovery, sentiment analysis, automatic summarization, question-answering, and fact-checking. However, automating this proc...

We introduce CLIMATE-FEVER, a new publicly available dataset for verification of climate change-related claims. By providing a dataset for the research community, we aim to facilitate and encourage work on improving algorithms for retrieving evidential support for climate-specific claims, addressing the underlying language understanding challenges,...

Climate change communication in the mass media and other textual sources may affect and shape public perception. Extracting climate change information from these sources is an important task, e.g., for filtering content and e-discovery, sentiment analysis, automatic summarization, question-answering, and fact-checking. However, automating this proc...

We present SUMO, a neural attention-based approach that learns to establish the correctness of textual claims based on evidence in the form of text documents (e.g., news articles or Web documents). SUMO further generates an extractive summary by presenting a diversified set of sentences from the documents that explain its decision on the correctnes...

We jointly explain the equity and value premium variations in a model with both short-run (SRR) and long-run (LRR) consumption risk. In our empirical analysis, we find that SRR varies with the business cycle, and it has a substantial predictive power for market excess returns and the value premium—both in-sample and out-of-sample. The LRR component...

We present SUMO, a neural attention-based approach that learns to establish the correctness of textual claims based on evidence in the form of text documents (e.g., news articles or Web documents). SUMO further generates an extractive summary by presenting a diversified set of sentences from the documents that explain its decision on the correctnes...

Measuring the congruence between two texts has several useful applications, such as detecting the prevalent deceptive and misleading news headlines on the web. Many works have proposed machine learning based solutions such as text similarity between the headline and body text to detect the incongruence. Text similarity based methods fail to perform...

There is ample evidence that factor momentum exists in the standard long–short mixed approach to factor investing. However, the excess returns are put under scrutiny due to the high implementation costs. We present a novel real‐life approach that relies on the long‐only integrated approach to factor investing. Instead of exploiting the potential mo...

Measuring congruence between two texts has several useful applications, such as detecting the prevalent deceptive and misleading news headlines on the web. Many works have proposed machine learning based solutions such as text similarity between the headline and body text to detect the incongruence. Text similarity based methods fail to perform wel...

Climate change may have a detrimental effect on a firm's financial performance. Using a forward-looking measure of climate risk exposure based on textual analysis of firms' 10-K reports, we assess whether climate risks---as disclosed to the regulator---are priced in the credit default swap (CDS) market. We construct this novel climate risk measure...

We propose a new method, VASA, based on variable subsample aggregation of model predictions for equity returns using a large-dimensional set of factors. To demonstrate the effectiveness, robustness, and dimension reduction power of VASA, we perform a comparative analysis between state-of-the-art machine learning algorithms. As a performance measure...

We jointly explain the variations of the equity and value premium in a model with both short-run (SRR) and long-run (LRR) consumption risk. In our preliminary empirical analysis, we find that SRR varies with the business cycle and it has a substantial predictive power for market excess returns and the value premium|both in-sample and out-of-sample....

This paper explores the performance of mixed-frequency predictive regressions for stock re- turns from the perspective of a Bayesian investor. We develop a parameter learning approach for sequential estimation, allowing for belief revisions. Empirically, we find that mixed-frequency models improve predictability, not only because of the combination...

A particle filter approach for general mixed-frequency state-space models is considered. It employs a backward smoother to filter high-frequency state variables from low-frequency observations. Moreover, it preserves the sequential nature of particle filters, allows for non-Gaussian shocks and nonlinear state-measurement relation, and alleviates th...

In this paper, we theoretically and empirically study the intrahorizon value at risk (iVaR) in a general jump-diffusion setting. We propose a new class of models of asset returns, the displaced mixed exponential model, which can arbitrarily closely approximate finite and infinite activity Lévy processes. We then derive analytical results for the iV...

To explore the rationality and competitiveness of the mutual fund industry, we analyze the alpha of active and index mutual funds from a global sample of more than 60,000 equity and fixed income funds and test the null hypothesis that alphas to investors are zero. We distinguish between institutional and retail investors since there are significant...

The concept of second-order risk operationalizes the estimation risk in portfolio construction induced by model uncertainty. We study its contribution to the realized volatility of recently developed risk parity strategies. For each strategy, we derive closed-form solutions for the second-order risk, subsequently illustrated in empirical analysis b...

There is ample evidence that factor momentum exists in the standard long--short mixed approach to factor investing. However, the excess returns are put under scrutiny due to the high implementation costs. We present a novel real-life approach that relies on the long-only integrated approach to factor investing. Instead of exploiting the potential m...

We estimate a flexible affine model using an unbalanced panel containing S&P 500 and VIX index returns and option prices and analyze the contribution of VIX options to the model's in- and out-of-sample performance. We find that they contain valuable information on the risk-neutral conditional distributions of volatility at different time horizons,...

We present a prediction model to forecast corporate defaults. In a theoretical model, under incomplete information in a market with publicly traded equity, we show that our approach must outperform ratings, Altman’s Z -score, and Merton’s distance to default. We reconcile the statistical and structural approaches under a common framework; that is,...

We study the difference between the returns to the integrated approach to style investing and those to the mixed approach. Unlike the mixed approach, the integrated approach aggregates factor characteristics at security level. Recent literature finds that the integrated approach dominates the mixed approach. Using statistical tools for robust perfo...

We study the impact of economic policy uncertainty on the term structure of nominal interest rates. We develop a general equilibrium model, in which both the government and the central bank policy decisions are driven by uncertainty shocks. Our affine yield curve model captures both the shape of the interest rate term structure as well as the hump-...

We investigate the implications of technological innovation and non-diversifiable risk on entrepreneurial entry and optimal portfolio choice. In a real options model where two risk-averse individuals strategically decide on technology adoption, we show that the impact of non-diversifiable risk on the option timing decision is ambiguous and depends...

Pursuing risk-based allocation across a universe of commodity assets, we find diversified risk parity (DRP) strategies to provide convincing results. DRP strives for maximum diversification along uncorrelated risk sources. A straightforward way to derive uncorrelated risk sources relies on principal components analysis (PCA). While the ensuing stat...

We analyze American put options in a hyper-exponential jump-diffusion model. Our contribution is threefold. Firstly, by following a maturity randomization approach, we solve the partial integro-differential equation and obtain a tight lower bound for the American option price. Secondly, our method allows to disentangle the contributions of jumps an...

We study the intra-horizon value at risk (iVaR) in a general jump diffusion setup and propose a new model of asset returns called displaced mixed-exponential model, which can arbitrarily closely approximate finite-activity jump-diffusions and completely monotone Levy processes. We derive analytical results for the iVaR and disentangle the risk cont...