arXiv subjects

Yu-Chien Tang

Publications and source records attributed to Yu-Chien Tang.

12 recordsLinked to original sources

arXiv2026

EgoArgus: Benchmarking VLMs as Situational Assistants for Modality-Grounded User Supports

VLMs are increasingly positioned as daily assistants that perceive first-person environments, follow user dialogue, and decide how to help. Existing egocentric benchmarks mainly evaluate visual understanding in isolation, leaving open whether models can arbitrate between visual evidence and user-provided language when the two are helpful, irrelevant, or conflicting. We introduce EgoArgus, a human-annotated dataset for evaluating egocentric assistants on understanding and decision tasks in five dialogue-video daily scenarios. Our results demonstrate that it is still challenging for current VLMs as reliable egocentric assistants, which requires identifying which modality is trustworthy and deciding when intervention is warranted. Deeper analysis also shows that existing modality bias mitigation methods are quite restricted to enhance performance, providing insights to aid practioners into the deployment of current VLMs as daily assistants.

cs.CL↗

arXiv2026

VDAR-Router: Adaptive LLMs Routing via Verbalized Query Difficulty Analysis Retrieval

Large language models are increasingly used in practical systems, making efficient model selection important for reducing deployment cost. LLM routing has emerged as a practical solution for allocating each input query to an appropriate model under a desired cost-performance trade-off. Existing routing methods often estimate model suitability from the surface semantics or embedding similarity of the input query. However, such methods may ignore the underlying difficulty of a query, leading to suboptimal routing decisions. To address the challenge, we propose VDAR-Router, a difficulty-aware retrieval-based routing framework. For each input query, VDAR-Router first generates an explicit difficulty analysis. It then retrieves historical examples with similar difficulty profiles. Based on the retrieved records, it estimates candidate model suitability and selects the model using a reward function that considers both performance and cost. Experiments on three datasets show that VDAR-Router consistently achieves better cost-performance trade-offs than existing baselines. These results demonstrate the effectiveness of difficulty-aware retrieval for training-free LLM routing. Case studies further show that explicit query analysis helps retrieve more relevant examples and supports more reliable routing decisions.

cs.CL↗

arXiv2026

VISTA: A Controllable Platform for Generating and Auditing Egocentric Assistance Scenarios

Evaluating whether AI agents can proactively assist humans in daily activities, ranging from routine household tasks to urgent safety-critical situations, requires diverse visual data. However, collecting such scenarios in the real world is often difficult, costly, or unsafe, and simulation environments often lack the social commonsense needed to simulate the consequences of different actions. In this work, we present VISTA, a controllable platform that uses a user-provided scenario seed, defined as a short natural-language description of the intended assistance situation, to generate editable plans, egocentric videos, and an auditable review trail. VISTA structures scenario intent around three interaction modes, including reactive, explicit proactive, and implicit proactive, and two consequence families, including safety-critical and everyday inconvenience, with no-assistance cases as controls. Its six-stage pipeline exposes the design brief, timed event script, first-frame plan, motion plan, and video plan, allowing users to revise each artifact in natural language before explicitly authorizing media generation. A human evaluation shows that videos retained by the complete VISTA workflow align more closely with their scenario seeds than outputs from two one-pass baselines. VISTA thereby makes targeted egocentric scenario generation inspectable, revisable, and empirically auditable.

cs.CL↗

arXiv2026

ConceptKT: A Benchmark for Concept-Level Deficiency Prediction in Knowledge Tracing

Knowledge Tracing (KT) is a critical technique for modeling student knowledge to support personalized learning. However, most KT systems focus on binary correctness prediction and cannot diagnose the underlying conceptual misunderstandings that lead to errors. Such fine-grained diagnostic feedback is essential for designing targeted instruction and effective remediation. In this work, we introduce the task of concept-level deficiency prediction, which extends traditional KT by identifying the specific concepts a student is likely to struggle with on future problems. We present ConceptKT, a dataset annotated with labels that capture both the concepts required to solve each question and the missing concepts underlying incorrect responses. We investigate in-context learning approaches to KT and evaluate the diagnostic capabilities of various Large Language Models (LLMs) and Large Reasoning Models (LRMs). Different strategies for selecting informative historical records are explored. Experimental results demonstrate that selecting response histories based on conceptual alignment and semantic similarity leads to improved performance on both correctness prediction and concept-level deficiency identification.

cs.CL↗

arXiv2026

OI-Bench: An Option Injection Benchmark for Evaluating LLM Susceptibility to Directive Interference

Benchmarking large language models (LLMs) is critical for understanding their capabilities, limitations, and robustness. In addition to interface artifacts, prior studies have shown that LLM decisions can be influenced by directive signals such as social cues, framing, and instructions. In this work, we introduce option injection, a benchmarking approach that augments the multiple-choice question answering (MCQA) interface with an additional option containing a misleading directive, leveraging standardized choice structure and scalable evaluation. We construct OI-Bench, a benchmark of 3,000 questions spanning knowledge, reasoning, and commonsense tasks, with 16 directive types covering social compliance, bonus framing, threat framing, and instructional interference. This setting combines manipulation of the choice interface with directive-based interference, enabling systematic assessment of model susceptibility. We evaluate 12 LLMs to analyze attack success rates, behavioral responses, and further investigate mitigation strategies ranging from inference-time prompting to post-training alignment. Experimental results reveal substantial vulnerabilities and heterogeneous robustness across models. OI-Bench is expected to support more systematic evaluation of LLM robustness to directive interference within choice-based interfaces.

cs.CL↗

arXiv2025

CARPAS: Towards Content-Aware Refinement of Provided Aspects for Summarization in Large Language Models

Aspect-based summarization has attracted significant attention for its ability to generate more fine-grained and user-aligned summaries. While most existing approaches assume a set of predefined aspects as input, real-world scenarios often present challenges where these given aspects may be incomplete, irrelevant, or entirely missing from the document. Users frequently expect systems to adaptively refine or filter the provided aspects based on the actual content. In this paper, we initiate this novel task setting, termed Content-Aware Refinement of Provided Aspects for Summarization (CARPAS), with the aim of dynamically adjusting the provided aspects based on the document context before summarizing. We construct three new datasets and conduct pilot experiments with four representative prompting strategies. Our analysis reveals that LLMs often over-generate aspects, resulting in excessively long and misaligned summaries. Building on this observation, we introduce a two-stage framework that derives lightweight scope guidance before aspect refinement and summarization, improving focus and reducing over-generation. Our extensive experiments show that the proposed approach significantly improves performance across all datasets. Moreover, our deeper analyses uncover LLMs' compliance when the requested number of aspects differs from their own estimations, establishing a crucial insight for the deployment of LLMs in similar real-world applications.

cs.CL↗

arXiv2025

MathEDU: Feedback Generation on Problem-Solving Processes for Mathematical Learning Support

The increasing reliance on Large Language Models (LLMs) across various domains extends to education, where students progressively use generative AI as a tool for learning. While prior work has examined LLMs' mathematical ability, their reliability in grading authentic student problem-solving processes and delivering effective feedback remains underexplored. This study introduces MathEDU, a dataset consisting of student problem-solving processes in mathematics and corresponding teacher-written feedback. We systematically evaluate the reliability of various models across three hierarchical tasks: answer correctness classification, error identification, and feedback generation. Experimental results show that fine-tuning strategies effectively improve performance in classifying correctness and locating erroneous steps. However, the generated feedback across models shows a considerable gap from teacher-written feedback. Critically, the generated feedback is often verbose and fails to provide targeted explanations for the student's underlying misconceptions. This emphasizes the urgent need for trustworthy and pedagogy-aware AI feedback in education.

cs.CL↗

arXiv2025

Stay Hungry, Stay Foolish: On the Extended Reading Articles Generation with LLMs

The process of creating educational materials is both time-consuming and demanding for educators. This research explores the potential of Large Language Models (LLMs) to streamline this task by automating the generation of extended reading materials and relevant course suggestions. Using the TED-Ed Dig Deeper sections as an initial exploration, we investigate how supplementary articles can be enriched with contextual knowledge and connected to additional learning resources. Our method begins by generating extended articles from video transcripts, leveraging LLMs to include historical insights, cultural examples, and illustrative anecdotes. A recommendation system employing semantic similarity ranking identifies related courses, followed by an LLM-based refinement process to enhance relevance. The final articles are tailored to seamlessly integrate these recommendations, ensuring they remain cohesive and informative. Experimental evaluations demonstrate that our model produces high-quality content and accurate course suggestions, assessed through metrics such as Hit Rate, semantic similarity, and coherence. Our experimental analysis highlight the nuanced differences between the generated and existing materials, underscoring the model's capacity to offer more engaging and accessible learning experiences. This study showcases how LLMs can bridge the gap between core content and supplementary learning, providing students with additional recommended resources while also assisting teachers in designing educational materials.

cs.CL↗

arXiv2025

Template-Based Financial Report Generation in Agentic and Decomposed Information Retrieval

Tailoring structured financial reports from companies' earnings releases is crucial for understanding financial performance and has been widely adopted in real-world analytics. However, existing summarization methods often generate broad, high-level summaries, which may lack the precision and detail required for financial reports that typically focus on specific, structured sections. While Large Language Models (LLMs) hold promise, generating reports adhering to predefined multi-section templates remains challenging. This paper investigates two LLM-based approaches popular in industry for generating templated financial reports: an agentic information retrieval (IR) framework and a decomposed IR approach, namely AgenticIR and DecomposedIR. The AgenticIR utilizes collaborative agents prompted with the full template. In contrast, the DecomposedIR approach applies a prompt chaining workflow to break down the template and reframe each section as a query answered by the LLM using the earnings release. To quantitatively assess the generated reports, we evaluated both methods in two scenarios: one using a financial dataset without direct human references, and another with a weather-domain dataset featuring expert-written reports. Experimental results show that while AgenticIR may excel in orchestrating tasks and generating concise reports through agent collaboration, DecomposedIR statistically significantly outperforms AgenticIR approach in providing broader and more detailed coverage in both scenarios, offering reflection on the utilization of the agentic framework in real-world applications.

cs.IR↗

arXiv2023

RSVP: Customer Intent Detection via Agent Response Contrastive and Generative Pre-Training

The dialogue systems in customer services have been developed with neural models to provide users with precise answers and round-the-clock support in task-oriented conversations by detecting customer intents based on their utterances. Existing intent detection approaches have highly relied on adaptively pre-training language models with large-scale datasets, yet the predominant cost of data collection may hinder their superiority. In addition, they neglect the information within the conversational responses of the agents, which have a lower collection cost, but are significant to customer intent as agents must tailor their replies based on the customers' intent. In this paper, we propose RSVP, a self-supervised framework dedicated to task-oriented dialogues, which utilizes agent responses for pre-training in a two-stage manner. Specifically, we introduce two pre-training tasks to incorporate the relations of utterance-response pairs: 1) Response Retrieval by selecting a correct response from a batch of candidates, and 2) Response Generation by mimicking agents to generate the response to a given utterance. Our benchmark results for two real-world customer service datasets show that RSVP significantly outperforms the state-of-the-art baselines by 4.95% for accuracy, 3.4% for MRR@3, and 2.75% for MRR@5 on average. Extensive case studies are investigated to show the validity of incorporating agent responses into the pre-training stage.

cs.CL↗

arXiv2023

NYCU-TWO at Memotion 3: Good Foundation, Good Teacher, then you have Good Meme Analysis

This paper presents a robust solution to the Memotion 3.0 Shared Task. The goal of this task is to classify the emotion and the corresponding intensity expressed by memes, which are usually in the form of images with short captions on social media. Understanding the multi-modal features of the given memes will be the key to solving the task. In this work, we use CLIP to extract aligned image-text features and propose a novel meme sentiment analysis framework, consisting of a Cooperative Teaching Model (CTM) for Task A and a Cascaded Emotion Classifier (CEC) for Tasks B&C. CTM is based on the idea of knowledge distillation, and can better predict the sentiment of a given meme in Task A; CEC can leverage the emotion intensity suggestion from the prediction of Task C to classify the emotion more precisely in Task B. Experiments show that we achieved the 2nd place ranking for both Task A and Task B and the 4th place ranking for Task C, with weighted F1-scores of 0.342, 0.784, and 0.535 respectively. The results show the robustness and effectiveness of our framework. Our code is released at github.

cs.CL↗

arXiv2020

EmotionGIF-Yankee: A Sentiment Classifier with Robust Model Based Ensemble Methods

This paper provides a method to classify sentiment with robust model based ensemble methods. We preprocess tweet data to enhance coverage of tokenizer. To reduce domain bias, we first train tweet dataset for pre-trained language model. Besides, each classifier has its strengths and weakness, we leverage different types of models with ensemble methods: average and power weighted sum. From the experiments, we show that our approach has achieved positive effect for sentiment classification. Our system reached third place among 26 teams from the evaluation in SocialNLP 2020 EmotionGIF competition.

cs.CL↗