Yue Zhang · page 2
Papers
- What Have We Achieved on Text Summarization? (2020 · 143 citations)
ROUGE gains on benchmarks such as CNN/DM say little about which errors summarizers actually make. This paper addresses that gap with a detailed manual error analysis. Drawing on the Multidimensional Quality Metric (MQM) framework, the authors define PolyTope, a set of 8 error categories split between Accuracy and… - MAGE: Machine-generated Text Detection in the Wild (2024 · 129 citations)
The paper asks whether AI-written text can still be detected in realistic conditions, where a detector does not know which domain a text comes from or which model produced it. - G-Transformer for Document-Level Machine Translation (2021 · 109 citations)
Bao et al. study why a standard Transformer often fails when an entire document is translated as one sequence. Under plain supervised training, long inputs, small datasets or large models can make BLEU collapse. The authors show this collapse is not overfitting. Instead, training gets trapped in local minima where… - AMR-to-text Generation with Synchronous Node Replacement Grammar (2017 · 109 citations)
The paper generates sentences directly from the AMR graph, without first making a spanning tree, so that no information is lost in the graph-to-tree step. - In-Order Transition-based Constituent Parsing (2017 · 99 citations)
Transition-based constituent parsers usually build trees in one of two orders. Bottom-up parsers (post-order traversal) can use features from subtrees they have already built, but they have no lookahead about structure still to come. Top-down parsers (pre-order traversal) give global guidance, but they must commit to… - SynGEC: Syntax-Enhanced Grammatical Error Correction with a Tailored GEC-Oriented Parser (2022 · 76 citations)
Grammatical error correction (GEC) systems usually treat input as a flat token sequence, even though many errors, such as tense or number agreement, depend on syntactic relations. - Transition-Based Syntactic Linearization (2015 · 52 citations)
Syntactic linearization takes a bag of words and optional syntactic constraints and builds the sentence and its tree together. - Universal Dependencies Parsing for Colloquial Singaporean English (2017 · 49 citations)
The authors ask why an English parser fails on Singlish, an English-based creole with imported Hokkien, Malay and Cantonese words, topic-first sentences, copula deletion and clause-final particles such as "lah". - Discriminative Syntax-Based Word Ordering for Text Generation (2015 · 48 citations)
The paper addresses word ordering for text generation: turning an unordered bag of words (or lemmas) into a fluent, grammatical sentence. - Exploiting Sentiment and Common Sense for Zero-shot Stance Detection (2022 · 30 citations)
Zero-shot stance detection asks a model to decide whether a text supports, opposes, or is neutral toward a topic it never saw in training. - NovelQA: Benchmarking Question Answering on Documents Exceeding 200K Tokens (2024 · 29 citations)
Current LLMs accept context windows of 200K+ tokens, but long-context benchmarks typically average around 60K tokens. - ZPR2: Joint Zero Pronoun Recovery and Resolution using Multi-Task Learning and BERT (2020 · 29 citations)
Chinese often omits pronouns, especially in conversation, which hurts machine translation, QA and dialogue systems. - How Likely Do LLMs with CoT Mimic Human Reasoning? (2025 · 22 citations)
Chain-of-thought prompting does not reliably improve accuracy, and the reasoning it produces may not reflect what drives a model's answer. - Personality Alignment of Large Language Models (2024 · 21 citations)
Standard LLM alignment targets broad, shared human values and ignores how individual users differ. The authors call this gap "Personality Alignment": adapting a model's answers and choices to one person or a small, closely related group. To make the task measurable, they build PAPI, a psychometrics-based dataset that… - Synergetic Event Understanding: A Collaborative Approach to Cross-Document Event Coreference Resolution with Large Language Models (2024 · 19 citations)
The paper addresses cross-document event coreference resolution (CDECR), the task of grouping event mentions from different documents that describe the same real-world event. - A Rationale-centric Counterfactual Data Augmentation Method for Cross-Document Event Coreference Resolution (2024 · 17 citations)
Cross-document event coreference resolution (ECR) groups event mentions from different documents that describe the same real-world event. - Constituency Parsing using LLMs (2023 · 14 citations)
The paper asks whether modern large language models can produce phrase-structure trees, a task where supervised parsers reach about 96% while unsupervised methods remain below 65%. - Assessing the Prosody of Non-Native Speakers of English: Measures and Feature Sets (2016 · 14 citations)
The paper addresses automatic rating of how natural the prosody (melody and rhythm) of non-native English speech sounds. - Lost in Literalism: How Supervised Training Shapes Translationese in LLMs (2025 · 12 citations)
LLMs learn from huge amounts of native text, yet their translations often still sound overly literal. The authors examine this "translationese" in LLM-based machine translation and trace it to supervised fine-tuning (SFT). Professional translators marked literal or unnatural spans in English-Chinese and… - Dual Cache for Long Document Neural Coreference Resolution (2023 · 12 citations)
Cache-based coreference models handle long documents by reading mentions left to right and keeping only a bounded set of entities in memory. - SR-LLM: Rethinking the Structured Representation in Large Language Model (2025 · 11 citations)
Structured representations such as AMR, parse syntax trees (PST) and first-order logic (FOL) have long been central to NLP. - Investigating Forgetting in Pre-Trained Representations Through Continual Learning (2023 · 10 citations)
When a pre-trained language model is fine-tuned on a sequence of tasks, its contextualized representations shift. - Cross-domain Generalization for AMR Parsing (2022 · 5 citations)
AMR parsers score above 84% SMATCH on their home test set, but performance outside that domain had barely been studied. - LLMEval-Fair: A Large-Scale Longitudinal Study on Robust and Fair Evaluation of Large Language Models (2026 · 3 citations)
Fixed public test sets can leak into training corpora and be overfitted, so leaderboard scores may overstate real LLM ability. - AiraXiv: An AI-Driven Open-Access Platform for Human and AI Scientists (2026 · 1 citation)
The authors argue that conference- and journal-based publishing struggles to keep up as LLMs and AI scientist systems sharply increase the number of manuscripts.
This page in Markdown → https://md.research.elamur.ai/free/en/authors/yue-zhang/page/2/