Julian Nyarko

Neel Guha, Julian Nyarko, Daniel E. Ho et al.

The advent of large language models (LLMs) and their adoption by the legal community has given rise to the question: what types of legal reasoning can LLMs perform? To enable greater study of this question, we present LegalBench: a collaboratively constructed legal reasoning benchmark consisting of 162 tasks covering six different types of legal reasoning. LegalBench was built through an interdisciplinary process, in which we collected tasks designed and hand-crafted by legal professionals. Because these subject matter experts took a leading role in construction, tasks either measure legal reasoning capabilities that are practically useful, or measure reasoning skills that lawyers find interesting. To enable cross-disciplinary conversations about LLMs in the law, we additionally show how popular legal frameworks for describing legal reasoning -- which distinguish between its many forms -- correspond to LegalBench tasks, thus giving lawyers and LLM developers a common vocabulary. This paper describes LegalBench, presents an empirical evaluation of 20 open-source and commercial LLMs, and illustrates the types of research explorations LegalBench enables.

Neel Guha, Daniel E. Ho, Julian Nyarko et al.

Can foundation models be guided to execute tasks involving legal reasoning? We believe that building a benchmark to answer this question will require sustained collaborative efforts between the computer science and legal communities. To that end, this short paper serves three purposes. First, we describe how IRAC-a framework legal scholars use to distinguish different types of legal reasoning-can guide the construction of a Foundation Model oriented benchmark. Second, we present a seed set of 44 tasks built according to this framework. We discuss initial findings, and highlight directions for new tasks. Finally-inspired by the Open Science movement-we make a call for the legal and computer science communities to join our efforts by contributing new tasks. This work is ongoing, and our progress can be tracked here: https://github.com/HazyResearch/legalbench.

Alex Chohlas-Wood, Madison Coots, Sharad Goel et al.

Predictive algorithms are now used to help distribute a large share of our society's resources and sanctions, such as healthcare, loans, criminal detentions, and tax audits. Under the right circumstances, these algorithms can improve the efficiency and equity of decision-making. At the same time, there is a danger that the algorithms themselves could entrench and exacerbate disparities, particularly along racial, ethnic, and gender lines. To help ensure their fairness, many researchers suggest that algorithms be subject to at least one of three constraints: (1) no use of legally protected features, such as race, ethnicity, and gender; (2) equal rates of "positive" decisions across groups; and (3) equal error rates across groups. Here we show that these constraints, while intuitively appealing, often worsen outcomes for individuals in marginalized groups, and can even leave all groups worse off. The inherent trade-off we identify between formal fairness constraints and welfare improvements -- particularly for the marginalized -- highlights the need for a more robust discussion on what it means for an algorithm to be "fair". We illustrate these ideas with examples from healthcare and the criminal-legal system, and make several proposals to help practitioners design more equitable algorithms.

Russell Yang, Ruishi Chen, Pierce Kelaita et al.

Two methodologies dominate current practices of benchmarking: rubric-based scoring evaluates items against predefined criteria, whereas comparative judgment elicits pairwise preferences between outputs. Although both methodologies are widely used, the choice between them is rarely justified. We release JudgmentBench, a benchmark of 30 real-world legal tasks, paired with 1,539 rubric scores and 1,530 pairwise preference judgments collected from practicing attorneys--including at major U.S. law firms--with substantial experience. The annotations constitute the first publicly available dataset in a high-expertise domain in which both supervision signals are elicited from the same experts on the same items. Using LLM-generated outputs at three constructed quality levels, we provide an initial empirical comparison: comparative judgments recover the intended quality ordering substantially better than rubrics (mean Spearman's rank correlation of 0.908 vs. 0.150, estimated difference = 0.758 [0.494, 1.021]) while requiring less than half the annotation time. The patterns hold for human annotators and LLM autograders. Beyond this initial comparison, the paired structure of the dataset supports a broader research agenda on how expert judgment should be elicited, aggregated, and used as supervision in domains without verifiable ground truth.

Alejandro Salinas, Amit Haim, Julian Nyarko

We employ an audit design to investigate biases in state-of-the-art large language models, including GPT-4. In our study, we prompt the models for advice involving a named individual across a variety of scenarios, such as during car purchase negotiations or election outcome predictions. We find that the advice systematically disadvantages names that are commonly associated with racial minorities and women. Names associated with Black women receive the least advantageous outcomes. The biases are consistent across 42 prompt templates and several models, indicating a systemic issue rather than isolated incidents. While providing numerical, decision-relevant anchors in the prompt can successfully counteract the biases, qualitative details have inconsistent effects and may even increase disparities. Our findings underscore the importance of conducting audits at the point of LLM deployment and implementation to mitigate their potential for harm against marginalized communities.

Sibo Ma, Alejandro Salinas, Peter Henderson et al.

We employ model pruning to examine how LLMs conceptualize racial biases, and whether a generalizable mitigation strategy for such biases appears feasible. Our analysis yields several novel insights. We find that pruning can be an effective method to reduce bias without significantly increasing anomalous model behavior. Neuron-based pruning strategies generally yield better results than approaches pruning entire attention heads. However, our results also show that the effectiveness of either approach quickly deteriorates as pruning strategies become more generalized. For instance, a model that is trained on removing racial biases in the context of financial decision-making poorly generalizes to biases in commercial transactions. Overall, our analysis suggests that racial biases are only partially represented as a general concept within language models. The other part of these biases is highly context-specific, suggesting that generalizable mitigation strategies may be of limited effectiveness. Our findings have important implications for legal frameworks surrounding AI. In particular, they suggest that an effective mitigation strategy should include the allocation of legal responsibility on those that deploy models in a specific use case.

Sibo Ma, Julian Nyarko

Data attribution seeks to trace model behavior back to the training examples that shaped it, enabling debugging, auditing, and data valuation at scale. Classical influence-function methods offer a principled foundation but remain impractical for modern networks because they require expensive backpropagation or Hessian inversion at inference. We propose a data attribution method that preserves the same first-order counterfactual target while eliminating per-query backward passes. Our approach simulates each training example's parameter influence through short-horizon gradient propagation during training and later reads out attributions for any query using only forward evaluations. This design shifts computation from inference to simulation, reflecting real deployment regimes where a model may serve billions of user queries but originate from a fixed, finite set of data sources (for example, a large language model trained on diverse corpora while compensating a specific publisher such as the New York Times). Empirically, on standard MLP benchmarks, our estimator matches or surpasses state-of-the-art baselines such as TRAK on standard attribution metrics (LOO and LDS) while offering orders-of-magnitude lower inference cost. By combining influence-function fidelity with first-order scalability, our method provides a theoretical framework for practical, real-time data attribution in large pretrained models.

Peter Adelson, Julian Nyarko

This paper introduces the Material Contracts Corpus (MCC), a publicly available dataset comprising over one million contracts filed by public companies with the U.S. Securities and Exchange Commission (SEC) between 2000 and 2023. The MCC facilitates empirical research on contract design and legal language, and supports the development of AI-based legal tools. Contracts in the corpus are categorized by agreement type and linked to specific parties using machine learning and natural language processing techniques, including a fine-tuned LLaMA-2 model for contract classification. The MCC further provides metadata such as filing form, document format, and amendment status. We document trends in contractual language, length, and complexity over time, and highlight the dominance of employment and security agreements in SEC filings. This resource is available for bulk download and online access at https://mcc.law.stanford.edu.

Sibo Ma, Julian Nyarko

We introduce a new method to identify emerging concepts in large text corpora. By analyzing changes in the heatmaps of the underlying embedding space, we are able to detect these concepts with high accuracy shortly after they originate, in turn outperforming common alternatives. We further demonstrate the utility of our approach by analyzing speeches in the U.S. Senate from 1941 to 2015. Our results suggest that the minority party is more active in introducing new concepts into the Senate discourse. We also identify specific concepts that closely correlate with the Senators' racial, ethnic, and gender identities. An implementation of our method is publicly available.

Rishi Bommasani, Drew A. Hudson, Ehsan Adeli et al.

AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.