arXiv subjects

Hangxiao Zhu

Publications and source records attributed to Hangxiao Zhu.

7 recordsLinked to original sources

arXiv2026

MemeBridge: A Dataset for Benchmarking and Mitigating the Bidirectional Cultural Gap in Meme Interpretation

Communicating across cultures is inherently challenging, especially through culturally dense and ambiguous formats like memes. While people expect large language models (LLMs) to hold promise for bridging such gaps, existing benchmark datasets often fail to capture the cultural context necessary for accurate interpretation. To address this, we introduce MemeBridge, a curated dataset centered on U.S.-originated memes, designed to capture two complementary perspectives: (1) how Chinese participants interpret these memes, and (2) how U.S. participants anticipate how people from other cultures might misunderstand them. Here, context refers to implicit cultural knowledge, including background beliefs, norms, and shared assumptions that shape meme comprehension. The dataset was constructed via a multi-stage crowdsourcing pipeline with rigorous validation, including human agreement checks and GPT-based classification verification. Each meme is annotated with sentiment, emotion, cultural significance, and knowledge type, providing rich supervision for downstream tasks. Notably, we observe that the anticipated misunderstandings from U.S. participants are often inaccurate, highlighting the asymmetries in cultural understanding and the challenges of adopting perspectives beyond one's own. This bidirectional framing, which focuses on both expression and perception, enables more nuanced benchmarking of cross-cultural comprehension. Our probing of multiple LLMs reveals that while models developed in different cultural contexts exhibit partial cross-cultural understanding, they often struggle with sophisticated interpretations. By contrast, fine-tuning with MemeBridge improves model performance, underscoring the value of culturally grounded resources for training and evaluating LLMs in globally diverse settings.

cs.CL↗

arXiv2026

Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction

Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic search, it becomes a bottleneck: exact lexical constraints, sparse clue conjunctions, local context checks, and multi-step hypothesis refinement are difficult to implement by calling a conventional off-the-shelf retriever, and evidence filtered out early cannot be recovered by stronger downstream reasoning. Agentic tasks further exacerbate this limitation because they require agents to orchestrate multiple steps, including discovering intermediate entities, combining weak clues, and revising the plan after observing partial evidence. To tackle the limitation, we study direct corpus interaction (DCI), where an agent searches the raw corpus directly with general-purpose terminal tools (e.g., grep, file reads, shell commands, lightweight scripts), without any embedding model, vector index, or retrieval API. This approach requires no offline indexing and adapts naturally to evolving local corpora. Across IR benchmarks and end-to-end agentic search tasks, this simple setup substantially outperforms strong sparse, dense, and reranking baselines on several BRIGHT and BEIR datasets, and attains strong accuracy on BrowseComp-Plus and multi-hop QA without relying on any conventional semantic retriever. Our results indicate that as language agents become stronger, retrieval quality depends not only on reasoning ability but also on the resolution of the interface through which the model interacts with the corpus, with which DCI opens a broader interface-design space for agentic search.

cs.IR↗

arXiv2026

SciImpact: A Multi-Dimensional, Multi-Field Benchmark for Scientific Impact Prediction

The rapid growth of scientific literature calls for automated methods to assess and predict research impact. Prior work has largely focused on citation-based metrics, leaving limited evaluation of models' capability to reason about other impact dimensions. To this end, we introduce SciImpact, a large-scale, multi-dimensional benchmark for scientific impact prediction spanning 19 fields. SciImpact captures various forms of scientific influence, ranging from citation counts to award recognition, media attention, patent reference, and artifact adoption, by integrating heterogeneous data sources and targeted web crawling. It comprises 215,928 contrastive paper pairs reflecting meaningful impact differences in both short-term (e.g., Best Paper Award) and long-term settings (e.g., Nobel Prize). We evaluate 11 widely used large language models (LLMs) on SciImpact. Results show that off-the-shelf models exhibit substantial variability across dimensions and fields, while multi-task supervised fine-tuning consistently enables smaller LLMs (e.g., 4B) to markedly outperform much larger models (e.g., 30B) and surpass powerful closed-source LLMs (e.g., o4-mini). These results establish SciImpact as a challenging benchmark and demonstrate its value for multi-dimensional, multi-field scientific impact prediction. Our project homepage is https://flypig23.github.io/sciimpact-homepage/

cs.CL↗

arXiv2025

Survivors, Complainers, and Borderliners: Upward Bias in Online Discussions of Academic Conference Reviews

Online discussion platforms, such as community Q&A sites and forums, have become important hubs where academic conference authors share and seek information about the peer review process and outcomes. However, these discussions involve only a subset of all submissions, raising concerns about the representativeness of the self-reported review scores. In this paper, we conduct a systematic study comparing the review score distributions of self-reported submissions in online discussions (based on data collected from Zhihu and Reddit) with those of all submissions. We reveal a consistent upward bias: the score distribution of self-reported samples is shifted upward relative to the population score distribution, with this difference statistically significant in most cases. Our analysis identifies three distinct contributors to this bias: (1) survivors, authors of accepted papers who are more likely to share good results than those of rejected papers who tend to conceal bad ones; (2) complainers, authors of high-scoring rejected papers who are more likely to voice complaints about the peer review process or outcomes than those of low scores; and (3) borderliners, authors with borderline scores who face greater uncertainty prior to decision announcements and are more likely to seek advice during the rebuttal period. These findings have important implications for how information seekers should interpret online discussions of academic conference reviews.

cs.SI↗

arXiv2025

TutorUp: What If Your Students Were Simulated? Training Tutors to Address Engagement Challenges in Online Learning

With the rise of online learning, many novice tutors lack experience engaging students remotely. We introduce TutorUp, a Large Language Model (LLM)-based system that enables novice tutors to practice engagement strategies with simulated students through scenario-based training. Based on a formative study involving two surveys (N1=86, N2=102) on student engagement challenges, we summarize scenarios that mimic real teaching situations. To enhance immersion and realism, we employ a prompting strategy that simulates dynamic online learning dialogues. TutorUp provides immediate and asynchronous feedback by referencing tutor-students online session dialogues and evidence-based teaching strategies from learning science literature. In a within-subject evaluation (N=16), participants rated TutorUp significantly higher than a baseline system without simulation capabilities regarding effectiveness and usability. Our findings suggest that TutorUp provides novice tutors with more effective training to learn and apply teaching strategies to address online student engagement challenges.

cs.HC↗

arXiv2025

From Text to Trust: Empowering AI-assisted Decision Making with Adaptive LLM-powered Analysis

AI-assisted decision making becomes increasingly prevalent, yet individuals often fail to utilize AI-based decision aids appropriately especially when the AI explanations are absent, potentially as they do not %understand reflect on AI's decision recommendations critically. Large language models (LLMs), with their exceptional conversational and analytical capabilities, present great opportunities to enhance AI-assisted decision making in the absence of AI explanations by providing natural-language-based analysis of AI's decision recommendation, e.g., how each feature of a decision making task might contribute to the AI recommendation. In this paper, via a randomized experiment, we first show that presenting LLM-powered analysis of each task feature, either sequentially or concurrently, does not significantly improve people's AI-assisted decision performance. To enable decision makers to better leverage LLM-powered analysis, we then propose an algorithmic framework to characterize the effects of LLM-powered analysis on human decisions and dynamically decide which analysis to present. Our evaluation with human subjects shows that this approach effectively improves decision makers' appropriate reliance on AI in AI-assisted decision making.

cs.HC↗

arXiv2023

Synthetic Data Generation with Large Language Models for Text Classification: Potential and Limitations

The collection and curation of high-quality training data is crucial for developing text classification models with superior performance, but it is often associated with significant costs and time investment. Researchers have recently explored using large language models (LLMs) to generate synthetic datasets as an alternative approach. However, the effectiveness of the LLM-generated synthetic data in supporting model training is inconsistent across different classification tasks. To better understand factors that moderate the effectiveness of the LLM-generated synthetic data, in this study, we look into how the performance of models trained on these synthetic data may vary with the subjectivity of classification. Our results indicate that subjectivity, at both the task level and instance level, is negatively associated with the performance of the model trained on synthetic data. We conclude by discussing the implications of our work on the potential and limitations of leveraging LLM for synthetic data generation.

cs.CL↗