Daniel Lee

Cynthia Dwork, Daniel Lee, Huijia Lin et al. · harvard

We identify and explore connections between the recent literature on multi-group fairness for prediction algorithms and the pseudorandomness notions of leakage-resilience and graph regularity. We frame our investigation using new variants of multicalibration based on statistical distance and closely related to the concept of outcome indistinguishability. Adopting this perspective leads us not only to new, more efficient algorithms for multicalibration, but also to our graph theoretic results and a proof of a novel hardcore lemma for real-valued functions.

Daniel Lee, Owen Queen, James Zou

Chain-of-thought traces from large reasoning models can span tens of thousands of tokens, yet we lack a vocabulary for describing their internal structure. Previous methods developed to analyze chain-of-thought traces are either too rigid or not expressive enough, failing to capture features across domains and models. To remedy this, we develop ReasonOps, an unsupervised, expressive method for annotating chain-of-thought traces, providing succinct universal operators. Using ReasonOps, we analyze 44,662 traces from 12 thinking LLMs spanning 6 families across 8 reasoning benchmarks and discover that they share a common compositional structure: 7 recurring reasoning operators -- discourse-level moves such as backtracking, inferring, and hypothesizing -- that emerge from unsupervised clustering of sentence-initial 3-token pivots. These operators appear across every model family and benchmark domain, confirmed by three independent LLM judges who classify held-out samples at 70 -76% accuracy. We analyze the structure of operators on easy vs. hard problems, revealing that reflective operators are more helpful on hard problems and harm performance on easy problems. Operator sequences are highly model-identifying: a classifier trained on operator distributions alone recovers the source model with macro-AUC, revealing that each model family has a distinctive reasoning fingerprint. Structural operator features predict within-problem answer correctness well above baselines. Classifiers built on these operators reach WP-AUC and on AIME specifically. ReasonOps further enables early quality estimation well before the trace completes: we predict at WP-AUC for only 50% of the trace. The ReasonOps pipeline is unsupervised and annotation-free, enabling deep insights into LLM reasoning traces as well as strong downstream results on model identification and correctness prediction.

Simone Conia, Min Li, Daniel Lee et al.

Recent work in Natural Language Processing and Computer Vision has been using textual information -- e.g., entity names and descriptions -- available in knowledge graphs to ground neural models to high-quality structured data. However, when it comes to non-English languages, the quantity and quality of textual information are comparatively scarce. To address this issue, we introduce the novel task of automatic Knowledge Graph Enhancement (KGE) and perform a thorough investigation on bridging the gap in both the quantity and quality of textual information between English and non-English languages. More specifically, we: i) bring to light the problem of increasing multilingual coverage and precision of entity names and descriptions in Wikidata; ii) demonstrate that state-of-the-art methods, namely, Machine Translation (MT), Web Search (WS), and Large Language Models (LLMs), struggle with this task; iii) present M-NTA, a novel unsupervised approach that combines MT, WS, and LLMs to generate high-quality textual information; and, iv) study the impact of increasing multilingual coverage and precision of non-English textual information in Entity Linking, Knowledge Graph Completion, and Question Answering. As part of our effort towards better multilingual knowledge graphs, we also introduce WikiKGE-10, the first human-curated benchmark to evaluate KGE approaches in 10 languages across 7 language families.

Revanth Gangi Reddy, Daniel Lee, Yi R. Fung et al.

Timely and comprehensive understanding of emerging events is crucial for effective decision-making; automating situation report generation can significantly reduce the time, effort, and cost for intelligence analysts. In this work, we identify intelligence analysts' practices and preferences for AI assistance in situation report generation to guide the design strategies for an effective, trust-building interface that aligns with their thought processes and needs. Next, we introduce SmartBook, an automated framework designed to generate situation reports from large volumes of news data, creating structured reports by automatically discovering event-related strategic questions. These reports include multiple hypotheses (claims), summarized and grounded to sources with factual evidence, to promote in-depth situation understanding. Our comprehensive evaluation of SmartBook, encompassing a user study alongside a content review with an editing study, reveals SmartBook's effectiveness in generating accurate and relevant situation reports. Qualitative evaluations indicate over 80% of questions probe for strategic information, and over 90% of summaries produce tactically useful content, being consistently favored over summaries from a large language model integrated with web search. The editing study reveals that minimal information is removed from the generated text (under 2.5%), suggesting that SmartBook provides analysts with a valuable foundation for situation reports

Ronak Pradeep, Daniel Lee, Ali Mousavi et al.

The rapid advancement of Large Language Models (LLMs) and conversational assistants necessitates dynamic, scalable, and configurable conversational datasets for training and evaluation. These datasets must accommodate diverse user interaction modes, including text and voice, each presenting unique modeling challenges. Knowledge Graphs (KGs), with their structured and evolving nature, offer an ideal foundation for current and precise knowledge. Although human-curated KG-based conversational datasets exist, they struggle to keep pace with the rapidly changing user information needs. We present ConvKGYarn, a scalable method for generating up-to-date and configurable conversational KGQA datasets. Qualitative psychometric analyses confirm our method can generate high-quality datasets rivaling a popular conversational KGQA dataset while offering it at scale and covering a wide range of human-interaction configurations. We showcase its utility by testing LLMs on diverse conversations - exploring model behavior on conversational KGQA sets with different configurations grounded in the same KG fact set. Our results highlight the ability of ConvKGYarn to improve KGQA foundations and evaluate parametric knowledge of LLMs, thus offering a robust solution to the constantly evolving landscape of conversational assistants.

Revanth Gangi Reddy, Ye Liu, Wenting Zhao et al.

Maintaining large-scale, multilingual codebases hinges on accurately localizing issues, which requires mapping natural-language error descriptions to the relevant functions that need to be modified. However, existing ranking approaches are often Python-centric and perform a single-pass search over the codebase. This work introduces SweRank+, a framework that couples SweRankMulti, a cross-lingual code ranking tool, with SweRankAgent, an agentic search setup, for iterative, multi-turn reasoning over the code repository. SweRankMulti comprises a code embedding retriever and a listwise LLM reranker, and is trained using a carefully curated large-scale issue localization dataset spanning multiple popular programming languages. SweRankAgent adopts an agentic search loop that moves beyond single-shot localization with a memory buffer to reason and accumulate relevant localization candidates over multiple turns. Our experiments on issue localization benchmarks spanning various languages demonstrate new state-of-the-art performance with SweRankMulti, while SweRankAgent further improves localization over single-pass ranking.

Jieun Han, Daniel Lee, Haneul Yoo et al.

Personalized learning has gained attention in English as a Foreign Language (EFL) education, where engagement and motivation play crucial roles in reading comprehension. We propose a novel approach to generating personalized English reading comprehension tests tailored to students' interests. We develop a structured content transcreation pipeline using OpenAI's gpt-4o, where we start with the RACE-C dataset, and generate new passages and multiple-choice reading comprehension questions that are linguistically similar to the original passages but semantically aligned with individual learners' interests. Our methodology integrates topic extraction, question classification based on Bloom's taxonomy, linguistic feature analysis, and content transcreation to enhance student engagement. We conduct a controlled experiment with EFL learners in South Korea to examine the impact of interest-aligned reading materials on comprehension and motivation. Our results show students learning with personalized reading passages demonstrate improved comprehension and motivation retention compared to those learning with non-personalized materials.

Zhengyuan Zhu, Daniel Lee, Hong Zhang et al.

Recent advancements in retrieval-augmented generation (RAG) have demonstrated impressive performance in the question-answering (QA) task. However, most previous works predominantly focus on text-based answers. While some studies address multimodal data, they still fall short in generating comprehensive multimodal answers, particularly for explaining concepts or providing step-by-step tutorials on how to accomplish specific goals. This capability is especially valuable for applications such as enterprise chatbots and settings such as customer service and educational systems, where the answers are sourced from multimodal data. In this paper, we introduce a simple and effective framework named MuRAR (Multimodal Retrieval and Answer Refinement). MuRAR enhances text-based answers by retrieving relevant multimodal data and refining the responses to create coherent multimodal answers. This framework can be easily extended to support multimodal answers in enterprise chatbots with minimal modifications. Human evaluation results indicate that multimodal answers generated by MuRAR are more useful and readable compared to plain text answers.

Yifan Wang, David Stevens, Pranay Shah et al.

The growing demand for AI training data has transformed data annotation into a global industry, but traditional approaches relying on human annotators are often time-consuming, labor-intensive, and prone to inconsistent quality. We propose the Model-in-the-Loop (MILO) framework, which integrates AI/ML models into the annotation process. Our research introduces a collaborative paradigm that leverages the strengths of both professional human annotators and large language models (LLMs). By employing LLMs as pre-annotation and real-time assistants, and judges on annotator responses, MILO enables effective interaction patterns between human annotators and LLMs. Three empirical studies on multimodal data annotation demonstrate MILO's efficacy in reducing handling time, improving data quality, and enhancing annotator experiences. We also introduce quality rubrics for flexible evaluation and fine-grained feedback on open-ended annotations. The MILO framework has implications for accelerating AI/ML development, reducing reliance on human annotation alone, and promoting better alignment between human and machine values.

Yuan Tian, Daniel Lee, Fei Wu et al.

Text-to-SQL models, which parse natural language (NL) questions to executable SQL queries, are increasingly adopted in real-world applications. However, deploying such models in the real world often requires adapting them to the highly specialized database schemas used in specific applications. We find that existing text-to-SQL models experience significant performance drops when applied to new schemas, primarily due to the lack of domain-specific data for fine-tuning. This data scarcity also limits the ability to effectively evaluate model performance in new domains. Continuously obtaining high-quality text-to-SQL data for evolving schemas is prohibitively expensive in real-world scenarios. To bridge this gap, we propose SQLsynth, a human-in-the-loop text-to-SQL data annotation system. SQLsynth streamlines the creation of high-quality text-to-SQL datasets through human-LLM collaboration in a structured workflow. A within-subjects user study comparing SQLsynth with manual annotation and ChatGPT shows that SQLsynth significantly accelerates text-to-SQL data annotation, reduces cognitive load, and produces datasets that are more accurate, natural, and diverse. Our code is available at https://github.com/magic-YuanTian/SQLsynth.

Nikhil Sharma, Zheng Zhang, Daniel Lee et al.

High-quality feedback is essential for effective human-AI interaction. It bridges knowledge gaps, corrects digressions, and shapes system behavior; both during interaction and throughout model development. Yet despite its importance, human feedback to AI is often infrequent and low quality. This gap motivates a critical examination of human feedback during interactions with AIs. To understand and overcome the challenges preventing users from giving high-quality feedback, we conducted two studies examining feedback dynamics between humans and conversational agents (CAs). Our formative study, through the lens of Grice's maxims, identified four Feedback Barriers -- Common Ground, Verifiability, Communication, and Informativeness -- that prevent high-quality feedback by users. Building on these findings, we derive three design desiderata and show that systems incorporating scaffolds aligned with these desiderata enabled users to provide higher-quality feedback. Finally, we detail a call for action to the broader AI community for advances in Large Language Models capabilities to overcome Feedback Barriers.

Revanth Gangi Reddy, Tanay Dixit, Jiaxin Qin et al.

Wikipedia, a vast and continuously consulted knowledge base, faces significant challenges in maintaining up-to-date content due to its reliance on manual human editors. Inspired by the vision of continuous knowledge acquisition in NELL and fueled by advances in LLM-based agents, this paper introduces WiNELL, an agentic framework for continuously updating Wikipedia articles. Our approach employs a multi-agent framework to aggregate online information, select new and important knowledge for a target entity in Wikipedia, and then generate precise edit suggestions for human review. Our fine-grained editing models, trained on Wikipedia's extensive history of human edits, enable incorporating updates in a manner consistent with human editing behavior. Our editor models outperform both open-source instruction-following baselines and closed-source LLMs (e.g., GPT-4o) in key information coverage and editing efficiency. End-to-end evaluation on high-activity Wikipedia pages demonstrates WiNELL's ability to identify and suggest timely factual updates. This opens up a promising research direction in LLM agents for automatically updating knowledge bases in a never-ending fashion.

Evan Wang, Simon Chess, Daniel Lee et al.

As neural theorem provers become increasingly agentic, the ability to interpret and act on compiler feedback is critical. However, existing Lean datasets consist almost exclusively of correct proofs, offering little supervision for understanding and repairing failures. We study Lean proof repair as a supervised learning problem: given an erroneous proof and compiler feedback, predict both a corrected proof and a natural-language diagnosis grounded in the same feedback. We introduce APRIL (Automated Proof Repair in Lean), a dataset of 260,000 supervised tuples pairing systematically generated proof failures with compiler diagnostics and aligned repair and explanation targets. Training language models on APRIL substantially improves repair accuracy and feedback-conditioned reasoning; in our single-shot repair evaluation setting, a finetuned 4B-parameter model outperforms the strongest open-source baseline. We view diagnostic-conditioned supervision as a complementary training signal for feedback-using provers. Our dataset is available at \href{https://huggingface.co/datasets/uw-math-ai/APRIL}{this link}.

Simone Conia, Daniel Lee, Min Li et al.

Translating text that contains entity names is a challenging task, as cultural-related references can vary significantly across languages. These variations may also be caused by transcreation, an adaptation process that entails more than transliteration and word-for-word translation. In this paper, we address the problem of cross-cultural translation on two fronts: (i) we introduce XC-Translate, the first large-scale, manually-created benchmark for machine translation that focuses on text that contains potentially culturally-nuanced entity names, and (ii) we propose KG-MT, a novel end-to-end method to integrate information from a multilingual knowledge graph into a neural machine translation model by leveraging a dense retrieval mechanism. Our experiments and analyses show that current machine translation systems and large language models still struggle to translate texts containing entity names, whereas KG-MT outperforms state-of-the-art approaches by a large margin, obtaining a 129% and 62% relative improvement compared to NLLB-200 and GPT-4, respectively.

Daniel George, Charles Yeh, Daniel Lee et al.

Frozen visual embeddings (e.g., CLIP, DINOv2/v3, SSCD) power retrieval and integrity systems, yet their use on face-containing data is constrained by unmeasured identity leakage and a lack of deployable mitigations. We take an attacker-aware view and contribute: (i) a benchmark of visual embeddings that reports open-set verification at low false-accept rates, a calibrated diffusion-based template inversion check, and face-context attribution with equal-area perturbations; and (ii) propose a one-shot linear projector that removes an estimated identity subspace while preserving the complementary space needed for utility, which for brevity we denote as the identity sanitization projection ISP. Across CelebA-20 and VGGFace2, we show that these encoders are robust under open-set linear probes, with CLIP exhibiting relatively higher leakage than DINOv2/v3 and SSCD, robust to template inversion, and are context-dominant. In addition, we show that ISP drives linear access to near-chance while retaining high non-biometric utility, and transfers across datasets with minor degradation. Our results establish the first attacker-calibrated facial privacy audit of non-FR encoders and demonstrate that linear subspace removal achieves strong privacy guarantees while preserving utility for visual search and retrieval.

Kanana LLM Team, Yunju Bak, Hojin Lee et al.

We introduce Kanana, a series of bilingual language models that demonstrate exceeding performance in Korean and competitive performance in English. The computational cost of Kanana is significantly lower than that of state-of-the-art models of similar size. The report details the techniques employed during pre-training to achieve compute-efficient yet competitive models, including high quality data filtering, staged pre-training, depth up-scaling, and pruning and distillation. Furthermore, the report outlines the methodologies utilized during the post-training of the Kanana models, encompassing supervised fine-tuning and preference optimization, aimed at enhancing their capability for seamless interaction with users. Lastly, the report elaborates on plausible approaches used for language model adaptation to specific scenarios, such as embedding, retrieval augmented generation, and function calling. The Kanana model series spans from 2.1B to 32.5B parameters with 2.1B models (base, instruct, embedding) publicly released to promote research on Korean language models.

Shinbok Lee, Gaeun Seo, Daniel Lee et al.

This study investigates language models' generative capabilities in tool-use dialogs. We categorize the models' outputs in tool-use dialogs into four distinct types: Tool Call, Answer Completion, Slot Question, and Relevance Detection, which serve as aspects for evaluation. We introduce FunctionChat-Bench, comprising 700 evaluation items and automated assessment programs. Using this benchmark, we evaluate several language models that support function calling. Our findings indicate that while language models may exhibit high accuracy in single-turn Tool Call scenarios, this does not necessarily translate to superior generative performance in multi-turn environments. We argue that the capabilities required for function calling extend beyond generating tool call messages; they must also effectively generate conversational messages that engage the user.

Sunghee Jung, Donghun Lee, Shinbok Lee et al.

Tool-Augmented Larage Language Models (TA-LLMs) have shown promise in real-world applications, but face challenges in handling incomplete queries and out-of-scope requests. While existing approaches rely mainly on Supervised Fine-Tuning with expert trajectories, we propose DiaTool-DPO, a novel method that enhances TA-LLM's dialogue capabilities through Direct Preference Optimization. We model TA-LLM interactions as a Markov Decision Process with 5 distinct dialogue states and categorize user queries into 3 types based on their state transition trajectories. We automatically construct paired trajectory datasets of correct and incorrect dialogue flows and introduce a specialized objective loss for dialogue control. Our comprehensive evaluation demonstrates that DiaTool-DPO approaches GPT-4o's performance (94.8% in information gathering, 91% in tool call rejection) with substantial improvements over baseline (44% and 9.6% respectively) while maintaining core functionality. Our approach opens new possibilities for developing TA-LLMs that can handle diverse real-world scenarios without requiring additional expert demonstrations or human labeling.

Daniel Lee, Nikhil Sharma, Donghoon Shin et al. · uw

Generative AI has made image creation more accessible, yet aligning outputs with nuanced creative intent remains challenging, particularly for non-experts. Existing tools often require users to externalize ideas through prompts or references, limiting fluid exploration. We introduce ThematicPlane, a system that enables users to navigate and manipulate high-level semantic concepts (e.g., mood, style, or narrative tone) within an interactive thematic design plane. This interface bridges the gap between tacit creative intent and system control. In our exploratory study (N=6), participants engaged in divergent and convergent creative modes, often embracing unexpected results as inspiration or iteration cues. While they grounded their exploration in familiar themes, differing expectations of how themes mapped to outputs revealed a need for more explainable controls. Overall, ThematicPlane fosters expressive, iterative workflows and highlights new directions for intuitive, semantics-driven interaction in generative design tools.

Priyan Vaithilingam, Munyeong Kim, Frida-Cecilia Acosta-Parenteau et al.

How do we update AI memory of user intent as intent changes? We consider how an AI interface may assist the integration of new information into a repository of natural language data. Inspired by software engineering concepts like impact analysis, we develop methods and a UI for managing semantic changes with non-local effects, which we call "semantic conflict resolution." The user commits new intent to a project -- makes a "semantic commit" -- and the AI helps the user detect and resolve semantic conflicts within a store of existing information representing their intent (an "intent specification"). We develop an interface, SemanticCommit, to better understand how users resolve conflicts when updating intent specifications such as Cursor Rules and game design documents. A knowledge graph-based RAG pipeline drives conflict detection, while LLMs assist in suggesting resolutions. We evaluate our technique on an initial benchmark. Then, we report a 12 user within-subjects study of SemanticCommit for two task domains -- game design documents, and AI agent memory in the style of ChatGPT memories -- where users integrated new information into an existing list. Half of our participants adopted a workflow of impact analysis, where they would first flag conflicts without AI revisions then resolve conflicts locally, despite having access to a global revision feature. We argue that AI agent interfaces, such as software IDEs like Cursor and Windsurf, should provide affordances for impact analysis and help users validate AI retrieval independently from generation. Our work speaks to how AI agent designers should think about updating memory as a process that involves human feedback and decision-making.

Xiaobin Shen, Daniel Lee, Sumit Ranjan et al. · cmu

Enterprise conversational AI systems are becoming increasingly popular to assist users in completing daily tasks such as those in marketing and customer management. However, new users often struggle to ask effective questions, especially in emerging systems with unfamiliar or evolving capabilities. This paper proposes a framework to enhance question suggestions in conversational enterprise AI systems by generating proactive, context-aware questions that try to address immediate user needs while improving feature discoverability. Our approach combines periodic user intent analysis at the population level with chat session-based question generation. We evaluate the framework using real-world data from the AI Assistant for Adobe Experience Platform (AEP), demonstrating the improved usefulness and system discoverability of the AI Assistant.

Geeyang Tay, Wentao Ma, Jaewon Lee et al.

Automatic speech recognition (ASR) systems have achieved near-human accuracy on curated benchmarks, yet still fail in real-world voice agents under conditions that current evaluations do not systematically cover. Without diagnostic tools that isolate specific failure factors, practitioners cannot anticipate which conditions, in which languages, will cause what degree of degradation. We introduce WildASR, a multilingual (four-language) diagnostic benchmark sourced entirely from real human speech that factorizes ASR robustness along three axes: environmental degradation, demographic shift, and linguistic diversity. Evaluating seven widely used ASR systems, we find severe and uneven performance degradation, and model robustness does not transfer across languages or conditions. Critically, models often hallucinate plausible but unspoken content under partial or degraded inputs, creating concrete safety risks for downstream agent behavior. Our results demonstrate that targeted, factor-isolated evaluation is essential for understanding and improving ASR reliability in production systems. Besides the benchmark itself, we also present three analytical tools that practitioners can use to guide deployment decisions.

Chanwoo Chun, Abdulkadir Canatar, SueYeon Chung et al.

The global dimensionality of a neural representation manifold provides rich insight into the computational process underlying both artificial and biological neural networks. However, all existing measures of global dimensionality are sensitive to the number of samples, i.e., the number of rows and columns of the sample matrix. We show that, in particular, the participation ratio of eigenvalues, a popular measure of global dimensionality, is highly biased with small sample sizes, and propose a bias-corrected estimator that is more accurate with finite samples and with noise. On synthetic data examples, we demonstrate that our estimator can recover the true known dimensionality. We apply our estimator to neural brain recordings, including calcium imaging, electrophysiological recordings, and fMRI data, and to the neural activations in a large language model and show our estimator is invariant to the sample size. Finally, our estimators can additionally be used to measure the local dimensionalities of curved neural manifolds by weighting the finite samples appropriately.

Donghoon Shin, Daniel Lee, Gary Hsieh et al. · uw

Poster designing can benefit from synchronous feedback from target audiences. However, gathering audiences with diverse perspectives and reconciling them on design edits can be challenging. Recent generative AI models present opportunities to simulate human-like interactions, but it is unclear how they may be used for feedback processes in design. We introduce PosterMate, a poster design assistant that facilitates collaboration by creating audience-driven persona agents constructed from marketing documents. PosterMate gathers feedback from each persona agent regarding poster components, and stimulates discussion with the help of a moderator to reach a conclusion. These agreed-upon edits can then be directly integrated into the poster design. Through our user study (N=12), we identified the potential of PosterMate to capture overlooked viewpoints, while serving as an effective prototyping tool. Additionally, our controlled online evaluation (N=100) revealed that the feedback from an individual persona agent is appropriate given its persona identity, and the discussion effectively synthesizes the different persona agents' perspectives.

Daniel Lee, Harsh Sharma, Jieun Han et al.

Translating knowledge-intensive and entity-rich text between English and Korean requires transcreation to preserve language-specific and cultural nuances beyond literal, phonetic or word-for-word conversion. We evaluate 13 models (LLMs and MT models) using automatic metrics and human assessment by bilingual annotators. Our findings show LLMs outperform traditional MT systems but struggle with entity translation requiring cultural adaptation. By constructing an error taxonomy, we identify incorrect responses and entity name errors as key issues, with performance varying by entity type and popularity level. This work exposes gaps in automatic evaluation metrics and hope to enable future work in completing culturally-nuanced machine translation.

Akash V. Maharaj, David Arbour, Daniel Lee et al.

Enterprise AI Assistants are increasingly deployed in domains where accuracy is paramount, making each erroneous output a potentially significant incident. This paper presents a comprehensive framework for monitoring, benchmarking, and continuously improving such complex, multi-component systems under active development by multiple teams. Our approach encompasses three key elements: (1) a hierarchical ``severity'' framework for incident detection that identifies and categorizes errors while attributing component-specific error rates, facilitating targeted improvements; (2) a scalable and principled methodology for benchmark construction, evaluation, and deployment, designed to accommodate multiple development teams, mitigate overfitting risks, and assess the downstream impact of system modifications; and (3) a continual improvement strategy leveraging multidimensional evaluation, enabling the identification and implementation of diverse enhancement opportunities. By adopting this holistic framework, organizations can systematically enhance the reliability and performance of their AI Assistants, ensuring their efficacy in critical enterprise environments. We conclude by discussing how this multifaceted evaluation approach opens avenues for various classes of enhancements, paving the way for more robust and trustworthy AI systems.

Aneesh Vathul, Daniel Lee, Sheryl Chen et al.

Large Language Models (LLMs) have demonstrated impressive capabilities on a broad array of NLP tasks, but their tendency to produce hallucinations$\unicode{x2013}$plausible-sounding but factually incorrect content$\unicode{x2013}$poses severe challenges in high-stakes domains. Existing hallucination detection methods either bear the computational cost of multiple inference passes or sacrifice accuracy for efficiency with single-pass approaches, neither of which is ideal in resource-constrained environments such as edge devices. We propose the Shannon Entropy Distribution Hallucination Detector (ShED-HD), a novel hallucination detection framework that bridges this gap by classifying sequence-level entropy patterns using a lightweight BiLSTM architecture with single-headed attention. In contrast to prior approaches, ShED-HD efficiently detects distinctive uncertainty patterns across entire output sequences, preserving contextual awareness. Through in-depth evaluation on three datasets (BioASQ, TriviaQA, and Jeopardy Questions), we show that ShED-HD significantly outperforms other computationally efficient approaches in the out-of-distribution setting, while achieving comparable performance in the in-distribution setting. ShED-HD facilitates hallucination detection that is low-cost, accurate, and generalizable, improving the credibility of content generated by LLMs in resource-constrained environments where trustworthy AI functionality is crucial.

Saloni Potdar, Daniel Lee, Omar Attia et al.

Question answering systems for knowledge graph (KGQA), answer factoid questions based on the data in the knowledge graph. KGQA systems are complex because the system has to understand the relations and entities in the knowledge-seeking natural language queries and map them to structured queries against the KG to answer them. In this paper, we introduce Chronos, a comprehensive evaluation framework for KGQA at industry scale. It is designed to evaluate such a multi-component system comprehensively, focusing on (1) end-to-end and component-level metrics, (2) scalable to diverse datasets and (3) a scalable approach to measure the performance of the system prior to release. In this paper, we discuss the unique challenges associated with evaluating KGQA systems at industry scale, review the design of Chronos, and how it addresses these challenges. We will demonstrate how it provides a base for data-driven decisions and discuss the challenges of using it to measure and improve a real-world KGQA system.

Zelin Zhou, Simone Conia, Daniel Lee et al.

Multilingual knowledge graphs (KGs) provide high-quality relational and textual information for various NLP applications, but they are often incomplete, especially in non-English languages. Previous research has shown that combining information from KGs in different languages aids either Knowledge Graph Completion (KGC), the task of predicting missing relations between entities, or Knowledge Graph Enhancement (KGE), the task of predicting missing textual information for entities. Although previous efforts have considered KGC and KGE as independent tasks, we hypothesize that they are interdependent and mutually beneficial. To this end, we introduce KG-TRICK, a novel sequence-to-sequence framework that unifies the tasks of textual and relational information completion for multilingual KGs. KG-TRICK demonstrates that: i) it is possible to unify the tasks of KGC and KGE into a single framework, and ii) combining textual information from multiple languages is beneficial to improve the completeness of a KG. As part of our contributions, we also introduce WikiKGE10++, the largest manually-curated benchmark for textual information completion of KGs, which features over 25,000 entities across 10 diverse languages.

Yeshwanth Cherapanamjeri, Daniel Lee

A large body of work in the statistics and computer science communities dating back to Huber (Huber, 1960) has led to statistically and computationally efficient outlier-robust estimators. Two particular outlier models have received significant attention: the adversarial and heavy-tailed models. While the former models outliers as the result of a malicious adversary manipulating the data, the latter relaxes distributional assumptions on the data allowing outliers to naturally occur as part of the data generating process. In the first setting, the goal is to develop estimators robust to the largest fraction of outliers while in the second, one seeks estimators to combat the loss of statistical efficiency, where the dependence on the failure probability is paramount. Despite these distinct motivations, the algorithmic approaches to both these settings have converged, prompting questions on the relationship between the models. In this paper, we investigate and provide a principled explanation for this phenomenon. First, we prove that any adversarially robust estimator is also resilient to heavy-tailed outliers for any statistical estimation problem with i.i.d data. As a corollary, optimal adversarially robust estimators for mean estimation, linear regression, and covariance estimation are also optimal heavy-tailed estimators. Conversely, for arguably the simplest high-dimensional estimation task of mean estimation, we construct heavy-tailed estimators whose application to the adversarial setting requires any black-box reduction to remove almost all the outliers in the data. Taken together, our results imply that heavy-tailed estimation is likely easier than adversarially robust estimation opening the door to novel algorithmic approaches for the heavy-tailed setting. Additionally, confidence intervals obtained for adversarially robust estimation also hold with high-probability.

Xiaoran Fan, Riley Simmons-Edler, Daewon Lee et al.

Perceiving obstacles and avoiding collisions is fundamental to the safe operation of a robot system, particularly when the robot must operate in highly dynamic human environments. Proximity detection using on-robot sensors can be used to avoid or mitigate impending collisions. However, existing proximity sensing methods are orientation and placement dependent, resulting in blind spots even with large numbers of sensors. In this paper, we introduce the phenomenon of the Leaky Surface Wave (LSW), a novel sensing modality, and present AuraSense, a proximity detection system using the LSW. AuraSense is the first system to realize no-dead-spot proximity sensing for robot arms. It requires only a single pair of piezoelectric transducers, and can easily be applied to off-the-shelf robots with minimal modifications. We further introduce a set of signal processing techniques and a lightweight neural network to address the unique challenges in using the LSW for proximity sensing. Finally, we demonstrate a prototype system consisting of a single piezoelectric element pair on a robot manipulator, which validates our design. We conducted several micro benchmark experiments and performed more than 2000 on-robot proximity detection trials with various potential robot arm materials, colliding objects, approach patterns, and robot movement patterns. AuraSense achieves 100% and 95.3% true positive proximity detection rates when the arm approaches static and mobile obstacles respectively, with a true negative rate over 99%, showing the real-world viability of this system.

Daniel Lee, Georgy Noarov, Mallesh Pai et al.

We introduce a simple but general online learning framework in which a learner plays against an adversary in a vector-valued game that changes every round. Even though the learner's objective is not convex-concave (and so the minimax theorem does not apply), we give a simple algorithm that can compete with the setting in which the adversary must announce their action first, with optimally diminishing regret. We demonstrate the power of our framework by using it to (re)derive optimal bounds and efficient algorithms across a variety of domains, ranging from multicalibration to a large set of no regret algorithms, to a variant of Blackwell's approachability theorem for polytopes with fast convergence rates. As a new application, we show how to ``(multi)calibeat'' an arbitrary collection of forecasters -- achieving an exponentially improved dependence on the number of models we are competing against, compared to prior work.

Daniel Lee, Chao-Chun Liang, Keh-Yih Su

We present a novel approach to answer the Chinese elementary school Social Study Multiple Choice questions. Although BERT has demonstrated excellent performance on Reading Comprehension tasks, it is found not good at handling some specific types of questions, such as Negation, All-of-the-above, and None-of-the-above. We thus propose a novel framework to cascade BERT with a Pre-Processor and an Answer-Selector modules to tackle the above challenges. Experimental results show the proposed approach effectively improves the performance of BERT, and thus demonstrate the feasibility of supplementing BERT with additional modules.

Daniel Lee, Gopi Krishnan Rajbahadur, Dayi Lin et al.

It is becoming increasingly difficult for game developers to manage the cost of developing a game, while meeting the high expectations of gamers. One way to balance the increasing gamer expectation and development stress is to build an active modding community around the game. There exist several examples of games with an extremely active and successful modding community, with the Minecraft game being one of the most notable ones. This paper reports on an empirical study of 1,114 popular and 1,114 unpopular Minecraft mods from the CurseForge mod distribution platform, one of the largest distribution platforms for Minecraft mods. We analyzed the relationship between 33 features across 5 dimensions of mod characteristics and the popularity of mods (i.e., mod category, mod documentation, environmental context of the mod, remuneration for the mod, and community contribution for the mod), to understand the characteristics of popular Minecraft mods. We firstly verify that the studied dimensions have significant explanatory power in distinguishing the popularity of the studied mods. Then we evaluated the contribution of each of the 33 features across the 5 dimensions. We observed that popular mods tend to have a high quality description and promote community contribution.

Daniel Lee, Dayi Lin, Cor-Paul Bezemer et al.

Game developers cannot always meet the growing and changing needs of the gaming community, due to the often already overloaded schedules of developers. So-called modders can potentially assist game developers with addressing gamers' needs. Modders are enthusiasts who provide modifications or completely new content for a game. By supporting modders, game developers can meet the rapidly growing and varying needs of their gamer base. Modders have the potential to play a role in extending the life expectancy of a game, thereby saving game developers time and money, and leading to a better overall gaming experience for their gamer base. In this paper, we empirically study the metadata of 9,521 mods of the 20 most-modded games on the Nexus Mods distribution platform. Our goal is to provide useful insights into the modding community of the Nexus Mods distribution platform from a quantitative perspective, and to provide researchers with a solid foundation for future exploration of game mods. In doing so, game developers can potentially reduce development time and cost due to the increased replayability of their games through mods. We find that providing official support for mods can be beneficial for the perceived quality of the mods of a game. In addition, mod users are willing to submit bug reports for a mod. However, they often fail to do this in a systematic manner using the bug reporting tool of the Nexus Mods platform, resulting in low-quality bug reports which are difficult to resolve. Based on our findings, we recommend that game developers who desire an active modding community for their own games provide the modding community with an officially-supported modding tool. In addition, we recommend that mod distribution platforms, such as Nexus Mods, improve their bug reporting system to receive higher quality bug reports.

Dongyub Lee, Byeongil Ko, Myeong Cheol Shin et al.

Detecting disfluencies in spontaneous speech is an important preprocessing step in natural language processing and speech recognition applications. Existing works for disfluency detection have focused on designing a single objective only for disfluency detection, while auxiliary objectives utilizing linguistic information of a word such as named entity or part-of-speech information can be effective. In this paper, we focus on detecting disfluencies on spoken transcripts and propose a method utilizing named entity recognition (NER) and part-of-speech (POS) as auxiliary sequence labeling (SL) tasks for disfluency detection. First, we investigate cases that utilizing linguistic information of a word can prevent mispredicting important words and can be helpful for the correct detection of disfluencies. Second, we show that training a disfluency detection model with auxiliary SL tasks can improve its F-score in disfluency detection. Then, we analyze which auxiliary SL tasks are influential depending on baseline models. Experimental results on the widely used English Switchboard dataset show that our method outperforms the previous state-of-the-art in disfluency detection.

Daniel Lee, Rakesh Verma, Avisha Das et al.

In this paper, we revisit the challenging problem of unsupervised single-document summarization and study the following aspects: Integer linear programming (ILP) based algorithms, Parameterized normalization of term and sentence scores, and Title-driven approaches for summarization. We describe a new framework, NewsSumm, that includes many existing and new approaches for summarization including ILP and title-driven approaches. NewsSumm's flexibility allows to combine different algorithms and sentence scoring schemes seamlessly. Our results combining sentence scoring with ILP and normalization are in contrast to previous work on this topic, showing the importance of a broader search for optimal parameters. We also show that the new title-driven reduction idea leads to improvement in performance for both unsupervised and supervised approaches considered.

Dongyub Lee, Myeongcheol Shin, Taesun Whang et al.

Text summarization refers to the process that generates a shorter form of text from the source document preserving salient information. Many existing works for text summarization are generally evaluated by using recall-oriented understudy for gisting evaluation (ROUGE) scores. However, as ROUGE scores are computed based on n-gram overlap, they do not reflect semantic meaning correspondences between generated and reference summaries. Because Korean is an agglutinative language that combines various morphemes into a word that express several meanings, ROUGE is not suitable for Korean summarization. In this paper, we propose evaluation metrics that reflect semantic meanings of a reference summary and the original document, Reference and Document Aware Semantic Score (RDASS). We then propose a method for improving the correlation of the metrics with human judgment. Evaluation results show that the correlation with human judgment is significantly higher for our evaluation metrics than for ROUGE scores.

Daniel Yang, Tarik Tosun, Ben Eisner et al.

We present a novel approach to robotic grasp planning using both a learned grasp proposal network and a learned 3D shape reconstruction network. Our system generates 6-DOF grasps from a single RGB-D image of the target object, which is provided as input to both networks. By using the geometric reconstruction to refine the the candidate grasp produced by the grasp proposal network, our system is able to accurately grasp both known and unknown objects, even when the grasp location on the object is not visible in the input image. This paper presents the network architectures, training procedures, and grasp refinement method that comprise our system. Experiments demonstrate the efficacy of our system at grasping both known and unknown objects (91% success rate in a physical robot environment, 84% success rate in a simulated environment). We additionally perform ablation studies that show the benefits of combining a learned grasp proposal with geometric reconstruction for grasping, and also show that our system outperforms several baselines in a grasping task.

Runnan Chen, Yuexin Ma, Nenglun Chen et al.

Marking anatomical landmarks in cephalometric radiography is a critical operation in cephalometric analysis. Automatically and accurately locating these landmarks is a challenging issue because different landmarks require different levels of resolution and semantics. Based on this observation, we propose a novel attentive feature pyramid fusion module (AFPF) to explicitly shape high-resolution and semantically enhanced fusion features to achieve significantly higher accuracy than existing deep learning-based methods. We also combine heat maps and offset maps to perform pixel-wise regression-voting to improve detection accuracy. By incorporating the AFPF and regression-voting, we develop an end-to-end deep learning framework that improves detection accuracy by 7%~11% for all the evaluation metrics over the state-of-the-art method. We present ablation studies to give more insights into different components of our method and demonstrate its generalization capability and stability for unseen data from diverse devices.

Rakesh Verma, Samaneh Karimi, Daniel Lee et al.

In a disaster situation, first responders need to quickly acquire situational awareness and prioritize response based on the need, resources available and impact. Can they do this based on digital media such as Twitter alone, or newswire alone, or some combination of the two? We examine this question in the context of the 2015 Nepal Earthquakes. Because newswire articles are longer, effective summaries can be helpful in saving time yet giving key content. We evaluate the effectiveness of several unsupervised summarization techniques in capturing key content. We propose a method to link tweets written by the public and newswire articles, so that we can compare their key characteristics: timeliness, whether tweets appear earlier than their corresponding news articles, and content. A novel idea is to view relevant tweets as a summary of the matching news article and evaluate these summaries. Whenever possible, we present both quantitative and qualitative evaluations. One of our main findings is that tweets and newswire articles provide complementary perspectives that form a holistic view of the disaster situation.

Riley Simmons-Edler, Ben Eisner, Daniel Yang et al.

A major challenge in reinforcement learning is exploration, when local dithering methods such as epsilon-greedy sampling are insufficient to solve a given task. Many recent methods have proposed to intrinsically motivate an agent to seek novel states, driving the agent to discover improved reward. However, while state-novelty exploration methods are suitable for tasks where novel observations correlate well with improved reward, they may not explore more efficiently than epsilon-greedy approaches in environments where the two are not well-correlated. In this paper, we distinguish between exploration tasks in which seeking novel states aids in finding new reward, and those where it does not, such as goal-conditioned tasks and escaping local reward maxima. We propose a new exploration objective, maximizing the reward prediction error (RPE) of a value function trained to predict extrinsic reward. We then propose a deep reinforcement learning method, QXplore, which exploits the temporal difference error of a Q-function to solve hard exploration tasks in high-dimensional MDPs. We demonstrate the exploration behavior of QXplore on several OpenAI Gym MuJoCo tasks and Atari games and observe that QXplore is comparable to or better than a baseline state-novelty method in all cases, outperforming the baseline on tasks where state novelty is not well-correlated with improved reward.

Tarik Tosun, Eric Mitchell, Ben Eisner et al.

We present a novel method enabling robots to quickly learn to manipulate objects by leveraging a motion planner to generate "expert" training trajectories from a small amount of human-labeled data. In contrast to the traditional sense-plan-act cycle, we propose a deep learning architecture and training regimen called PtPNet that can estimate effective end-effector trajectories for manipulation directly from a single RGB-D image of an object. Additionally, we present a data collection and augmentation pipeline that enables the automatic generation of large numbers (millions) of training image and trajectory examples with almost no human labeling effort. We demonstrate our approach in a non-prehensile tool-based manipulation task, specifically picking up shoes with a hook. In hardware experiments, PtPNet generates motion plans (open-loop trajectories) that reliably (89% success over 189 trials) pick up four very different shoes from a range of positions and orientations, and reliably picks up a shoe it has never seen before. Compared with a traditional sense-plan-act paradigm, our system has the advantages of operating on sparse information (single RGB-D frame), producing high-quality trajectories much faster than the "expert" planner (300ms versus several seconds), and generalizing effectively to previously unseen shoes.

Riley Simmons-Edler, Ben Eisner, Eric Mitchell et al.

Off-Policy reinforcement learning (RL) is an important class of methods for many problem domains, such as robotics, where the cost of collecting data is high and on-policy methods are consequently intractable. Standard methods for applying Q-learning to continuous-valued action domains involve iteratively sampling the Q-function to find a good action (e.g. via hill-climbing), or by learning a policy network at the same time as the Q-function (e.g. DDPG). Both approaches make tradeoffs between stability, speed, and accuracy. We propose a novel approach, called Cross-Entropy Guided Policies, or CGP, that draws inspiration from both classes of techniques. CGP aims to combine the stability and performance of iterative sampling policies with the low computational cost of a policy network. Our approach trains the Q-function using iterative sampling with the Cross-Entropy Method (CEM), while training a policy network to imitate CEM's sampling behavior. We demonstrate that our method is more stable to train than state of the art policy network methods, while preserving equivalent inference time compute costs, and achieving competitive total reward on standard benchmarks.

Luis Moraes, Shahryar Baki, Rakesh Verma et al.

The CL-SciSumm 2016 shared task introduced an interesting problem: given a document D and a piece of text that cites D, how do we identify the text spans of D being referenced by the piece of text? The shared task provided the first annotated dataset for studying this problem. We present an analysis of our continued work in improving our system's performance on this task. We demonstrate how topic models and word embeddings can be used to surpass the previously best performing system.

Rakesh Verma, Daniel Lee

Due to its promise to alleviate information overload, text summarization has attracted the attention of many researchers. However, it has remained a serious challenge. Here, we first prove empirical limits on the recall (and F1-scores) of extractive summarizers on the DUC datasets under ROUGE evaluation for both the single-document and multi-document summarization tasks. Next we define the concept of compressibility of a document and present a new model of summarization, which generalizes existing models in the literature and integrates several dimensions of the summarization, viz., abstractive versus extractive, single versus multi-document, and syntactic versus semantic. Finally, we examine some new and existing single-document summarization algorithms in a single framework and compare with state of the art summarizers on DUC data.