arXiv subjects

Hannah Kim

Publications and source records attributed to Hannah Kim.

31 records · Page 2Linked to original sources

arXiv2024

A Blueprint Architecture of Compound AI Systems for Enterprise

Large Language Models (LLMs) have showcased remarkable capabilities surpassing conventional NLP challenges, creating opportunities for use in production use cases. Towards this goal, there is a notable shift to building compound AI systems, wherein LLMs are integrated into an expansive software infrastructure with many components like models, retrievers, databases and tools. In this paper, we introduce a blueprint architecture for compound AI systems to operate in enterprise settings cost-effectively and feasibly. Our proposed architecture aims for seamless integration with existing compute and data infrastructure, with ``stream'' serving as the key orchestration concept to coordinate data and instructions among agents and other components. Task and data planners, respectively, break down, map, and optimize tasks and data to available agents and data sources defined in respective registries, given production constraints such as accuracy and latency.

cs.DB↗

arXiv2024

MEGAnno+: A Human-LLM Collaborative Annotation System

Large language models (LLMs) can label data faster and cheaper than humans for various NLP tasks. Despite their prowess, LLMs may fall short in understanding of complex, sociocultural, or domain-specific context, potentially leading to incorrect annotations. Therefore, we advocate a collaborative approach where humans and LLMs work together to produce reliable and high-quality labels. We present MEGAnno+, a human-LLM collaborative annotation system that offers effective LLM agent and annotation management, convenient and robust LLM annotation, and exploratory verification of LLM labels by humans.

cs.CL↗

arXiv2024

Knowledge Acquisition and Integration with Expert-in-the-loop

Constructing and serving knowledge graphs (KGs) is an iterative and human-centered process involving on-demand programming and analysis. In this paper, we present Kyurem, a programmable and interactive widget library that facilitates human-in-the-loop knowledge acquisition and integration to enable continuous curation a knowledge graph (KG). Kyurem provides a seamless environment within computational notebooks where data scientists explore a KG to identify opportunities for acquiring new knowledge and verify recommendations provided by AI agents for integrating the acquired knowledge in the KG. We refined Kyurem through participatory design and conducted case studies in a real-world setting for evaluation. The case-studies show that introduction of Kyurem within an existing HR knowledge graph construction and serving platform improved the user experience of the experts and helped eradicate inefficiencies related to knowledge acquisition and integration tasks

cs.HC↗

arXiv2024

Characterizing Large Language Models as Rationalizers of Knowledge-intensive Tasks

Large language models (LLMs) are proficient at generating fluent text with minimal task-specific supervision. Yet, their ability to provide well-grounded rationalizations for knowledge-intensive tasks remains under-explored. Such tasks, like commonsense multiple-choice questions, require rationales based on world knowledge to support predictions and refute alternate options. We consider the task of generating knowledge-guided rationalization in natural language by using expert-written examples in a few-shot manner. Surprisingly, crowd-workers preferred knowledge-grounded rationales over crowdsourced rationalizations, citing their factuality, sufficiency, and comprehensive refutations. Although LLMs-generated rationales were preferable, further improvements in conciseness and novelty are required. In another study, we show how rationalization of incorrect model predictions erodes humans' trust in LLM-generated rationales. Motivated by these observations, we create a two-stage pipeline to review task predictions and eliminate potential incorrect decisions before rationalization, enabling trustworthy rationale generation.

cs.CL↗

arXiv2023

SemARFlow: Injecting Semantics into Unsupervised Optical Flow Estimation for Autonomous Driving

Unsupervised optical flow estimation is especially hard near occlusions and motion boundaries and in low-texture regions. We show that additional information such as semantics and domain knowledge can help better constrain this problem. We introduce SemARFlow, an unsupervised optical flow network designed for autonomous driving data that takes estimated semantic segmentation masks as additional inputs. This additional information is injected into the encoder and into a learned upsampler that refines the flow output. In addition, a simple yet effective semantic augmentation module provides self-supervision when learning flow and its boundaries for vehicles, poles, and sky. Together, these injections of semantic information improve the KITTI-2015 optical flow test error rate from 11.80% to 8.38%. We also show visible improvements around object boundaries as well as a greater ability to generalize across datasets. Code is available at https://github.com/duke-vision/semantic-unsup-flow-release.

cs.CV↗

arXiv2023

Real-time Likelihood Methods for Improved Gamma-ray Transient Detection and Localization

We present a maximum likelihood (ML) algorithm that is fast enough to detect gamma-ray transients in real time on low-performance processors often used for space applications. We validate the routine with simulations and find that, relative to algorithms based on excess counts, the ML method is nearly twice as sensitive, allowing detection of 240-280% more short gamma-ray bursts. We characterize a reference implementation of the code, estimating its computational complexity and benchmarking it on a range of processors. We exercise the reference implementation on archival data from the Fermi Gamma-ray Burst Monitor (GBM), verifying the sensitivity improvements. In particular, we show that the ML algorithm would have detected GRB 170817A even if it had been nearly four times fainter. We present an ad hoc but effective scheme for discriminating transients associated with background variations. We show that the on-board localizations generated by ML are accurate, but that refined off-line localizations require a detector response matrix with about ten times finer resolution than is current practice. Increasing the resolution of the GBM response matrix could substantially reduce the few-degree systematic uncertainty observed in the localizations of bright bursts.

astro-ph.HE↗

arXiv2023

Towards Transparent, Reusable, and Customizable Data Science in Computational Notebooks

Data science workflows are human-centered processes involving on-demand programming and analysis. While programmable and interactive interfaces such as widgets embedded within computational notebooks are suitable for these workflows, they lack robust state management capabilities and do not support user-defined customization of the interactive components. The absence of such capabilities hinders workflow reusability and transparency while limiting the scope of exploration of the end-users. In response, we developed MAGNETON, a framework for authoring interactive widgets within computational notebooks that enables transparent, reusable, and customizable data science workflows. The framework enhances existing widgets to support fine-grained interaction history management, reusable states, and user-defined customizations. We conducted three case studies in a real-world knowledge graph construction and serving platform to evaluate the effectiveness of these widgets. Based on the observations, we discuss future implications of employing MAGNETON widgets for general-purpose data science workflows.

cs.HC↗

arXiv2023

Towards Multifaceted Human-Centered AI

Human-centered AI workflows involve stakeholders with multiple roles interacting with each other and automated agents to accomplish diverse tasks. In this paper, we call for a holistic view when designing support mechanisms, such as interaction paradigms, interfaces, and systems, for these multifaceted workflows.

cs.DB↗

arXiv2023

MEGAnno: Exploratory Labeling for NLP in Computational Notebooks

We present MEGAnno, a novel exploratory annotation framework designed for NLP researchers and practitioners. Unlike existing labeling tools that focus on data labeling only, our framework aims to support a broader, iterative ML workflow including data exploration and model development. With MEGAnno's API, users can programmatically explore the data through sophisticated search and automated suggestion functions and incrementally update task schema as their project evolve. Combined with our widget, the users can interactively sort, filter, and assign labels to multiple items simultaneously in the same notebook where the rest of the NLP project resides. We demonstrate MEGAnno's flexible, exploratory, efficient, and seamless labeling experience through a sentiment analysis use case.

cs.HC↗

arXiv2022

Optical Flow Training under Limited Label Budget via Active Learning

Supervised training of optical flow predictors generally yields better accuracy than unsupervised training. However, the improved performance comes at an often high annotation cost. Semi-supervised training trades off accuracy against annotation cost. We use a simple yet effective semi-supervised training method to show that even a small fraction of labels can improve flow accuracy by a significant margin over unsupervised training. In addition, we propose active learning methods based on simple heuristics to further reduce the number of labels required to achieve the same target accuracy. Our experiments on both synthetic and real optical flow datasets show that our semi-supervised networks generally need around 50% of the labels to achieve close to full-label accuracy, and only around 20% with active learning on Sintel. We also analyze and show insights on the factors that may influence active learning performance. Code is available at https://github.com/duke-vision/optical-flow-active-learning-release.

cs.CV↗

arXiv2021

Out-of-Distribution Detection in Dermatology using Input Perturbation and Subset Scanning

Recent advances in deep learning have led to breakthroughs in the development of automated skin disease classification. As we observe an increasing interest in these models in the dermatology space, it is crucial to address aspects such as the robustness towards input data distribution shifts. Current skin disease models could make incorrect inferences for test samples from different hardware devices and clinical settings or unknown disease samples, which are out-of-distribution (OOD) from the training samples. To this end, we propose a simple yet effective approach that detect these OOD samples prior to making any decision. The detection is performed via scanning in the latent space representation (e.g., activations of the inner layers of any pre-trained skin disease classifier). The input samples could also perturbed to maximise divergence of OOD samples. We validate our ODD detection approach in two use cases: 1) identify samples collected from different protocols, and 2) detect samples from unknown disease classes. Additionally, we evaluate the performance of the proposed approach and compare it with other state-of-the-art methods. Furthermore, data-driven dermatology applications may deepen the disparity in clinical care across racial and ethnic groups since most datasets are reported to suffer from bias in skin tone distribution. Therefore, we also evaluate the fairness of these OOD detection methods across different skin tones. Our experiments resulted in competitive performance across multiple datasets in detecting OOD samples, which could be used (in the future) to design more effective transfer learning techniques prior to inferring on these samples.

cs.CV↗

arXiv2019

TopicSifter: Interactive Search Space Reduction Through Targeted Topic Modeling

Topic modeling is commonly used to analyze and understand large document collections. However, in practice, users want to focus on specific aspects or "targets" rather than the entire corpus. For example, given a large collection of documents, users may want only a smaller subset which more closely aligns with their interests, tasks, and domains. In particular, our paper focuses on large-scale document retrieval with high recall where any missed relevant documents can be critical. A simple keyword matching search is generally not effective nor efficient as 1) it is difficult to find a list of keyword queries that can cover the documents of interest before exploring the dataset, 2) some documents may not contain the exact keywords of interest but may still be highly relevant, and 3) some words have multiple meanings, which would result in irrelevant documents included in the retrieved subset. In this paper, we present TopicSifter, a visual analytics system for interactive search space reduction. Our system utilizes targeted topic modeling based on nonnegative matrix factorization and allows users to give relevance feedback in order to refine their target and guide the topic modeling to the most relevant results.

cs.IR↗

arXiv2018

Understanding Actors and Evaluating Personae with Gaussian Embeddings

Understanding narrative content has become an increasingly popular topic. Nonetheless, research on identifying common types of narrative characters, or personae, is impeded by the lack of automatic and broad-coverage evaluation methods. We argue that computationally modeling actors provides benefits, including novel evaluation mechanisms for personae. Specifically, we propose two actor-modeling tasks, cast prediction and versatility ranking, which can capture complementary aspects of the relation between actors and the characters they portray. For an actor model, we present a technique for embedding actors, movies, character roles, genres, and descriptive keywords as Gaussian distributions and translation vectors, where the Gaussian variance corresponds to actors' versatility. Empirical results indicate that (1) the technique considerably outperforms TransE (Bordes et al. 2013) and ablation baselines and (2) automatically identified persona topics (Bamman, O'Connor, and Smith 2013) yield statistically significant improvements in both tasks, whereas simplistic persona descriptors including age and gender perform inconsistently, validating prior research.

cs.CY↗