Chris Tanner

Rik Koncel-Kedziorski, Michael Krumdick, Viet Lai et al.

Answering questions within business and finance requires reasoning, precision, and a wide-breadth of technical knowledge. Together, these requirements make this domain difficult for large language models (LLMs). We introduce BizBench, a benchmark for evaluating models' ability to reason about realistic financial problems. BizBench comprises eight quantitative reasoning tasks, focusing on question-answering (QA) over financial data via program synthesis. We include three financially-themed code-generation tasks from newly collected and augmented QA data. Additionally, we isolate the reasoning capabilities required for financial QA: reading comprehension of financial text and tables for extracting intermediate values, and understanding financial concepts and formulas needed to calculate complex solutions. Collectively, these tasks evaluate a model's financial background knowledge, ability to parse financial documents, and capacity to solve problems with code. We conduct an in-depth evaluation of open-source and commercial LLMs, comparing and contrasting the behavior of code-focused and language-focused models. We demonstrate that the current bottleneck in performance is due to LLMs' limited business and financial understanding, highlighting the value of a challenging benchmark for quantitative reasoning within this domain.

Craig W. Schmidt, Chris Tanner, Yuval Pinter

Byte Pair Encoding (BPE) is a widely used tokenization algorithm, whose tokens cannot extend across pre-tokenization boundaries, functionally limiting it to representing at most full words. The BoundlessBPE and SuperBPE algorithms extend and improve BPE by relaxing this limitation and allowing the formation of superwords, which are combinations of pretokens that form phrases. However, previous implementations were impractical to train: for example, BoundlessBPE took 4.7 CPU days to train on 1GB of data. We show that supermerge candidates, two or more consecutive pretokens eligible to form a supermerge, can be aggregated by frequency much like regular pretokens. This avoids keeping full documents in memory, as the original implementations of BoundlessBPE and SuperBPE required, leading to a significant training speedup. We present a two-phase formulation of BoundlessBPE that separates first-phase learning of regular merges from second-phase learning of supermerges, producing identical results to the original implementation. We also show a near-equivalence between two-phase BoundlessBPE and SuperBPE, with the difference being that a manually selected hyperparameter used in SuperBPE can be automatically determined in the second phase of BoundlessBPE. These changes enable a much faster implementation, allowing training on that same 1GB of data in 603 and 593 seconds for BoundlessBPE and SuperBPE, respectively, a more than 600x increase in speed. For each of BoundlessBPE, SuperBPE, and BPE, we open-source both a reference Python implementation and a fast Rust implementation.

Ian Kelk, Benjamin Basseri, Wee Yi Lee et al.

Automatic fake news detection models are ostensibly based on logic, where the truth of a claim made in a headline can be determined by supporting or refuting evidence found in a resulting web query. These models are believed to be reasoning in some way; however, it has been shown that these same results, or better, can be achieved without considering the claim at all -- only the evidence. This implies that other signals are contained within the examined evidence, and could be based on manipulable factors such as emotion, sentiment, or part-of-speech (POS) frequencies, which are vulnerable to adversarial inputs. We neutralize some of these signals through multiple forms of both neural and non-neural pre-processing and style transfer, and find that this flattening of extraneous indicators can induce the models to actually require both claims and evidence to perform well. We conclude with the construction of a model using emotion vectors built off a lexicon and passed through an "emotional attention" mechanism to appropriately weight certain emotions. We provide quantifiable results that prove our hypothesis that manipulable features are being used for fact-checking.

Sahithya Ravi, Chris Tanner, Raymond Ng et al.

Event coreference models cluster event mentions pertaining to the same real-world event. Recent models rely on contextualized representations to recognize coreference among lexically or contextually similar mentions. However, models typically fail to leverage commonsense inferences, which is particularly limiting for resolving lexically-divergent mentions. We propose a model that extends event mentions with temporal commonsense inferences. Given a complex sentence with multiple events, e.g., "The man killed his wife and got arrested", with the target event "arrested", our model generates plausible events that happen before the target event - such as "the police arrived", and after it, such as "he was sentenced". We show that incorporating such inferences into an existing event coreference model improves its performance, and we analyze the coreferences in which such temporal knowledge is required.

Jilin Wang, Michael Krumdick, Baojia Tong et al.

Document layout analysis (DLA) is the task of detecting the distinct, semantic content within a document and correctly classifying these items into an appropriate category (e.g., text, title, figure). DLA pipelines enable users to convert documents into structured machine-readable formats that can then be used for many useful downstream tasks. Most existing state-of-the-art (SOTA) DLA models represent documents as images, discarding the rich metadata available in electronically generated PDFs. Directly leveraging this metadata, we represent each PDF page as a structured graph and frame the DLA problem as a graph segmentation and classification problem. We introduce the Graph-based Layout Analysis Model (GLAM), a lightweight graph neural network competitive with SOTA models on two challenging DLA datasets - while being an order of magnitude smaller than existing models. In particular, the 4-million parameter GLAM model outperforms the leading 140M+ parameter computer vision-based model on 5 of the 11 classes on the DocLayNet dataset. A simple ensemble of these two models achieves a new state-of-the-art on DocLayNet, increasing mAP from 76.8 to 80.8. Overall, GLAM is over 5 times more efficient than SOTA models, making GLAM a favorable engineering choice for DLA tasks.

Varshini Reddy, Craig W. Schmidt, Seth Ebner et al.

Large language models (LLMs) have achieved impressive proficiency in basic arithmetic, rivaling human-level performance on standard numerical tasks. However, little attention has been given to how these models perform when numerical expressions deviate from the prevailing conventions present in their training corpora. In this work, we investigate numerical reasoning across a wide range of numeral scripts and formats. We show that LLM accuracy drops substantially when numerical inputs are rendered in underrepresented scripts or formats, despite the underlying mathematical reasoning being identical. We further demonstrate that targeted prompting strategies, such as few-shot prompting and explicit numeral mapping, can greatly narrow this gap. Our findings highlight an overlooked challenge in multilingual numerical reasoning and provide actionable insights for working with LLMs to reliably interpret, manipulate, and generate numbers across diverse numeral scripts and formatting styles.

Anita Mahinpei, Zona Kostic, Chris Tanner

Data visualization captions help readers understand the purpose of a visualization and are crucial for individuals with visual impairments. The prevalence of poor figure captions and the successful application of deep learning approaches to image captioning motivate the use of similar techniques for automated figure captioning. However, research in this field has been stunted by the lack of suitable datasets. We introduce LineCap, a novel figure captioning dataset of 3,528 figures, and we provide insights from curating this dataset and using end-to-end deep learning models for automated figure captioning.

Xiaohan Yang, Eduardo Peynetti, Vasco Meerman et al.

Coreference resolution -- which is a crucial task for understanding discourse and language at large -- has yet to witness widespread benefits from large language models (LLMs). Moreover, coreference resolution systems largely rely on supervised labels, which are highly expensive and difficult to annotate, thus making it ripe for prompt engineering. In this paper, we introduce a QA-based prompt-engineering method and discern \textit{generative}, pre-trained LLMs' abilities and limitations toward the task of coreference resolution. Our experiments show that GPT-2 and GPT-Neo can return valid answers, but that their capabilities to identify coreferent mentions are limited and prompt-sensitive, leading to inconsistent results.

Craig W. Schmidt, Michael Krumdick, Adam Wiemerslage et al.

We introduce Tokenization with Split Trees (ToaST), a subword tokenization method that directly optimizes compression under a new recursive inference procedure. ToaST greedily splits each pretoken into a full binary tree using precomputed byte n-gram counts, independent of any vocabulary. Given a vocabulary, inference recursively descends each split tree and emits the first in-vocabulary node reached on each path. Vocabulary selection is formulated as an Integer Program (IP) that minimizes the total token count over all split trees under this inference procedure. The Linear Programming (LP) relaxation is near-integral in practice, yielding provably near-optimal vocabularies, with training time empirically scaling quadratically in the number of split trees. On English text, ToaST reduces token counts by more than 11% compared to BPE, WordPiece, and UnigramLM at vocabulary sizes of 40,960 and above, reducing the number of inference tokens for models using this tokenizer, thus extending the effective context length. ToaST also uses common single-byte tokens less frequently than these baselines, leading to a substantial improvement in Renyi efficiency. In experiments training 1.5B parameter language models, ToaST achieves the highest CORE score, outperforming baselines by 2.6%--7.6%, with significance for two of three, and scoring best on 13 of 22 individual tasks.

Blake Bullwinkel, Kristen Grabarz, Lily Ke et al.

Differentially private (DP) synthetic data is a promising approach to maximizing the utility of data containing sensitive information. Due to the suppression of underrepresented classes that is often required to achieve privacy, however, it may be in conflict with fairness. We evaluate four DP synthesizers and present empirical results indicating that three of these models frequently degrade fairness outcomes on downstream binary classification tasks. We draw a connection between fairness and the proportion of minority groups present in the generated synthetic data, and find that training synthesizers on data that are pre-processed via a multi-label undersampling method can promote more fair outcomes without degrading accuracy.

Michael Krumdick, Adam Wiemerslage, Seth Ebner et al.

Thousands of diverse benchmarks have been developed to measure the quality of large language models (LLMs). Yet prior work has demonstrated that LLM performance is often sufficiently explained by a small set of latent factors, or abilities. This suggests the potential for more efficient and principled benchmarking, but it remains difficult to compare the quality of different methods. Motivated by predictive validity, we argue that the quality of a benchmarking framework should be grounded in how efficiently it enables the prediction of model performance on unseen tasks. To analyze this objective, we collect the "Wide-scale Item Level Dataset" (WILD), a dataset of item-model response pairs, comprising evaluations of 65 models on 109,564 unique items spanning 163 tasks drawn from 27 datasets. This dataset enables the first analysis of how different techniques can predict a model's performance on a large, diverse collection of unseen tasks under different budget constraints. We demonstrate that combining a modified multidimensional item response theory (IRT) model with adaptive item selection driven by optimal experimental design can predict performance on 112 held-out benchmark tasks with a mean absolute error (MAE) of less than 7%, and can do so after observing only 16 items. We further demonstrate that incorporating cost-aware discount factors into our selection criteria can reduce the total tokens needed to reach 7% MAE from 141,000 tokens to only 22,000, an 85% reduction in evaluation cost.

Charles J. Lovering, Seth Ebner, Brandon Smock et al.

Professionals in academia, law, and finance audit their documents because inconsistencies can result in monetary, reputational, and scientific costs. Language models (LMs) have the potential to dramatically speed up this auditing process. To understand their abilities, we introduce a benchmark, FIND (Finding INconsistencies in Documents), where each example is a document with an inconsistency inserted manually by a domain expert. Despite the documents being long, technical, and complex, the best-performing model (gpt-5) recovered 64% of the inserted inconsistencies. Surprisingly, gpt-5 also found undiscovered inconsistencies present in the original documents. For example, on 50 arXiv papers, we judged 136 out of 196 of the model's suggestions to be legitimate inconsistencies missed by the original authors. However, despite these findings, even the best models miss almost half of the inconsistencies in FIND, demonstrating that inconsistency detection is still a challenging task.

Craig W. Schmidt, Varshini Reddy, Haoran Zhang et al.

Tokenization is a foundational step in natural language processing (NLP) tasks, bridging raw text and language models. Existing tokenization approaches like Byte-Pair Encoding (BPE) originate from the field of data compression, and it has been suggested that the effectiveness of BPE stems from its ability to condense text into a relatively small number of tokens. We test the hypothesis that fewer tokens lead to better downstream performance by introducing PathPiece, a new tokenizer that segments a document's text into the minimum number of tokens for a given vocabulary. Through extensive experimentation we find this hypothesis not to be the case, casting doubt on the understanding of the reasons for effective tokenization. To examine which other factors play a role, we evaluate design decisions across all three phases of tokenization: pre-tokenization, vocabulary construction, and segmentation, offering new insights into the design of effective tokenizers. Specifically, we illustrate the importance of pre-tokenization and the benefits of using BPE to initialize vocabulary construction. We train 64 language models with varying tokenization, ranging in size from 350M to 2.4B parameters, all of which are made publicly available.

Varshini Reddy, Rik Koncel-Kedziorski, Viet Dac Lai et al.

For large language models (LLMs) to be effective in the financial domain -- where each decision can have a significant impact -- it is necessary to investigate realistic tasks and data. Financial professionals often interact with documents that are hundreds of pages long, but most financial research datasets only deal with short excerpts from these documents. To address this, we introduce a long-document financial QA task. We augment 7,437 questions from the existing FinQA dataset with the full-document context, extending the average context length from under 700 words in FinQA to 123k words in DocFinQA. We conduct extensive experiments over retrieval-based QA pipelines and long-context language models. DocFinQA proves a significant challenge for even state-of-the-art systems. We also provide a case-study on the longest documents in DocFinQA and find that models particularly struggle on these documents. Addressing these challenges may have a wide reaching impact across applications where specificity and long-range contexts are critical, like gene sequences and legal document contract analysis.

Michael Krumdick, Charles Lovering, Varshini Reddy et al.

LLM-as-a-Judge is a framework that uses an LLM (large language model) to evaluate the quality of natural language text - typically text that is also generated by an LLM. This framework holds great promise due to its relative low-cost, ease of use, and strong correlations with human stylistic preferences. However, LLM Judges have been shown to exhibit biases that can distort their judgments. We evaluate how well LLM Judges can grade whether a given response to a conversational question is correct, an ability crucial to soundly estimating the overall response quality. To do so, we create and publicly release a human-annotated dataset with labels of correctness for 1,200 LLM responses. We source questions from a combination of existing datasets and a novel, challenging benchmark (BFF-Bench) created for this analysis. We demonstrate a strong connection between an LLM's ability to correctly answer a question and grade responses to that question. Although aggregate level statistics might imply a judge has high agreement with human annotators, it will struggle on the subset of questions it could not answer. To address this issue, we recommend a simple solution: provide the judge with a correct, human-written reference answer. We perform an in-depth analysis on how reference quality can affect the performance of an LLM Judge. We show that providing a weaker judge (e.g. Qwen 2.5 7B) with higher quality references reaches better agreement with human annotators than a stronger judge (e.g. GPT-4o) with synthetic references.

Omri Uzan, Craig W. Schmidt, Chris Tanner et al.

While subword tokenizers such as BPE and WordPiece are typically used to build vocabularies for NLP models, the method of decoding text into a sequence of tokens from these vocabularies is often left unspecified, or ill-suited to the method in which they were constructed. We provide a controlled analysis of seven tokenizer inference methods across four different algorithms and three vocabulary sizes, performed on a novel intrinsic evaluation suite we curated for English, combining measures rooted in morphology, cognition, and information theory. We show that for the most commonly used tokenizers, greedy inference performs surprisingly well; and that SaGe, a recently-introduced contextually-informed tokenizer, outperforms all others on morphological alignment.

Craig W. Schmidt, Varshini Reddy, Chris Tanner et al.

Pre-tokenization, the initial step in many modern tokenization pipelines, segments text into smaller units called pretokens, typically splitting on whitespace and punctuation. While this process encourages having full, individual words as tokens, it introduces a fundamental limitation in most tokenization algorithms such as Byte Pair Encoding (BPE). Specifically, pre-tokenization causes the distribution of tokens in a corpus to heavily skew towards common, full-length words. This skewed distribution limits the benefits of expanding to larger vocabularies, since the additional tokens appear with progressively lower counts. To overcome this barrier, we propose BoundlessBPE, a modified BPE algorithm that relaxes the pretoken boundary constraint. Our approach selectively merges two complete pretokens into a larger unit we term a superword. Superwords are not necessarily semantically cohesive. For example, the pretokens " of" and " the" might be combined to form the superword " of the". This merging strategy results in a substantially more uniform distribution of tokens across a corpus than standard BPE, and compresses text more effectively, with up to a 15% increase in bytes per token.

Varshini Reddy, Craig W. Schmidt, Yuval Pinter et al.

Tokenization, a crucial initial step in natural language processing, is governed by several key parameters, such as the tokenization algorithm, vocabulary size, pre-tokenization strategy, inference strategy, and training data corpus. This paper investigates the impact of an often-overlooked hyperparameter, tokenizer training data size. We train BPE, UnigramLM, and WordPiece tokenizers across various vocabulary sizes using English training data ranging from 1GB to 900GB. Our findings reveal diminishing returns as training data size increases beyond roughly 150GB, suggesting a practical limit to the improvements in tokenization quality achievable through additional data. We analyze this phenomenon and attribute the saturation effect to constraints introduced by the pre-tokenization stage. We then demonstrate the extent to which these findings can generalize by experimenting on data in Russian, a language typologically distant from English. For Russian text, we observe diminishing returns after training a tokenizer from 200GB of data, which is approximately 33% more than when training on English. These results provide valuable insights for optimizing the tokenization process by reducing the compute required for training on large corpora and suggest promising directions for future research in tokenization algorithms.

Charles Lovering, Michael Krumdick, Viet Dac Lai et al.

Some statements have one well-defined continuation (e.g., "the Eiffel Tower is in [Paris]"), whereas others have a natural distribution over multiple options (e.g., "the weighted coin flip was [Heads/Tails].") We argue that language model (LM) outputs should capture these natural distributions. Our work specifically tests whether LM output probabilities are calibrated to numeric information within their textual contexts. For example, if the context (the prompt) concerns two equally likely options (e.g., heads or tails for a fair coin), the LM output probabilities should also be equal. Likewise, in a context with nonuniformly likely events (e.g., rolling a pair with two dice) an LM should output proportionate probabilities. However, we find that even in simple settings, the best LMs (1) are poorly calibrated and (2) have systematic biases: artifacts like word identity, word order, and word frequency all impact calibration. For example, gpt-4o-mini often picks the first of two options presented in the prompt regardless of the options' implied likelihoods, whereas Llama-3.1-8B picks the second. Models do not allocate probability mass among valid options in a calibrated manner.

Michael Krumdick, Varshini Reddy, Shivani Chaudhary et al.

As concerns surrounding AI-driven labor displacement intensify in knowledge-intensive sectors, existing benchmarks fail to measure performance on tasks that define practical professional expertise. Finance, in particular, has been identified as a domain with high AI exposure risk, yet lacks robust benchmarks to track real-world developments. This gap is compounded by the absence of clear accountability mechanisms in current Large Language Model (LLM) deployments. To address this, we introduce FrontierFinance, a long-horizon benchmark of 25 complex financial modeling tasks across five core finance models, requiring an average of over 18 hours of skilled human labor per task to complete. Developed with financial professionals, the benchmark reflects industry-standard financial modeling workflows and is paired with detailed rubrics for structured evaluation. We engage human experts to define the tasks, create rubrics, grade LLMs, and perform the tasks themselves as human baselines. We demonstrate that our human experts both receive higher scores on average, and are more likely to provide client-ready outputs than current state-of-the-art systems.

Yifan Hu, Frank Liang, Dachuan Zhao et al.

Byte-Pair Encoding (BPE) has become a widely adopted subword tokenization method in modern language models due to its simplicity and strong empirical performance across downstream tasks. However, applying BPE to unsegmented languages such as Chinese presents significant challenges, as its frequency-driven merge operation is agnostic to linguistic boundaries. To address this, we propose two entropy-informed pre-tokenization strategies that guide BPE segmentation using unsupervised information-theoretic cues. The first approach uses pointwise mutual information and left/right entropy to identify coherent character spans, while the second leverages predictive entropy derived from a pretrained GPT-2 model to detect boundary uncertainty. We evaluate both methods on a subset of the PKU dataset and demonstrate substantial improvements in segmentation precision, recall, and F1 score compared to standard BPE. Our results suggest that entropy-guided pre-tokenization not only enhances alignment with gold-standard linguistic units but also offers a promising direction for improving tokenization quality in low-resource and multilingual settings.

Viet Dac Lai, Michael Krumdick, Charles Lovering et al.

The financial domain frequently deals with large numbers of long documents that are essential for daily operations. Significant effort is put towards automating financial data analysis. However, a persistent challenge, not limited to the finance domain, is the scarcity of datasets that accurately reflect real-world tasks for model evaluation. Existing datasets are often constrained by size, context, or relevance to practical applications. Moreover, LLMs are currently trained on trillions of tokens of text, limiting access to novel data or documents that models have not encountered during training for unbiased evaluation. We propose SEC-QA, a continuous dataset generation framework with two key features: 1) the semi-automatic generation of Question-Answer (QA) pairs spanning multiple long context financial documents, which better represent real-world financial scenarios; 2) the ability to continually refresh the dataset using the most recent public document collections, not yet ingested by LLMs. Our experiments show that current retrieval augmented generation methods systematically fail to answer these challenging multi-document questions. In response, we introduce a QA system based on program-of-thought that improves the ability to perform complex information retrieval and quantitative reasoning pipelines, thereby increasing QA accuracy.

Felix Grezes, Sergi Blanco-Cuaresma, Alberto Accomazzi et al.

The existing search tools for exploring the NASA Astrophysics Data System (ADS) can be quite rich and empowering (e.g., similar and trending operators), but researchers are not yet allowed to fully leverage semantic search. For example, a query for "results from the Planck mission" should be able to distinguish between all the various meanings of Planck (person, mission, constant, institutions and more) without further clarification from the user. At ADS, we are applying modern machine learning and natural language processing techniques to our dataset of recent astronomy publications to train astroBERT, a deeply contextual language model based on research at Google. Using astroBERT, we aim to enrich the ADS dataset and improve its discoverability, and in particular we are developing our own named entity recognition tool. We present here our preliminary results and lessons learned.