Papers by Ce Zhang
Multilingual Language Models Predict Human Reading Behavior (2021.naacl-main)
Copied to clipboard
| Challenge: | Recent studies show that cognitively motivated "attention" mechanism in neural models is not a good indicator for relative importance. |
| Approach: | They compare the performance of language-specific and multilingual pretrained transformer models to predict reading time measures reflecting natural human sentence processing. |
| Outcome: | The proposed models predict reading time measures on Dutch, English, German, and Russian texts. |
ZuCo 2.0: A Dataset of Physiological Recordings During Natural Reading and Annotation (2020.lrec-1)
Copied to clipboard
| Challenge: | a new dataset of eye-tracking and electroencephalography captures language understanding . eye movement data provides millisecond-accurate records of where humans look when reading . |
| Approach: | They recorded and preprocessed eye-tracking and electroencephalography data during natural reading and during annotation. |
| Outcome: | The study combines eye-tracking and electroencephalography to capture the reading process . the data can be used to evaluate state-of-the-art machine learning systems . |
Dynamic Human Evaluation for Relative Model Comparisons (2022.lrec-1)
Copied to clipboard
| Challenge: | Automated metrics have reported flaws when applied to measure quality aspects of generated text and have been shown to correlate poorly with human judgements. |
| Approach: | They propose an agent-based framework to measure the required number of human annotations when evaluating generated outputs in relative comparison settings. |
| Outcome: | The proposed model can be compared with a crowdsourced case study and a simulation with simulated human judgements. |
Mitigating Catastrophic Forgetting in Language Transfer via Model Merging (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models have shown remarkable capabilities, particularly in English, but for less prevalent languages, performance can be significantly lower, making additional adaptation paramount. |
| Approach: | They propose a new adaptation method based on iteratively merging multiple models fine-tuned on a subset of available training data that reduces forgetting while maintaining learning on the target domain. |
| Outcome: | The proposed method outperforms LLAMA-3-8B-based models in German and German while maintaining learning on the target domain. |
Entity Recognition at First Sight: Improving NER with Eye Movement Information (N19-1)
Copied to clipboard
| Challenge: | Previous studies have shown eye-tracking data can be used to improve natural language processing models. |
| Approach: | They leverage eye movement features from three corpora with recorded gaze information to augment a neural model for named entity recognition with gaze embeddings. |
| Outcome: | The proposed model outperforms baseline models on both individual datasets and in cross-domain settings. |
WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models (2026.acl-long)
Copied to clipboard
Rui Wang, Ce Zhang, Jun-Yu Ma, Jianshu Zhang, Hongru Wang, Yi Chen, Boyang Xue, Tianqing Fang, Zhisong Zhang, Hongming Zhang, Haitao Mi, Dong Yu, Kam-Fai Wong
| Challenge: | Existing agentic systems are retrieval-heavy but reasoning-light . current systems lack compositional reasoning, a key component of deep research . |
| Approach: | They propose a data synthesis pipeline WebAggregator to shift agentic paradigm . they use Proactive Explorer to collect interconnected knowledge and Compositional Logic Proposer to weave knowledge into complex questions . |
| Outcome: | The proposed pipeline surpasses GPT-4.1 and matches Claude-3.7-Sonnet on GAIA, WebWalkerQA, and XBench. |
Towards Automatic Bias Detection in Knowledge Graphs (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies have shown that knowledge graphs are prone to various social biases, and have proposed multiple methods for debiasing them. |
| Approach: | They propose a framework for identifying biases present in knowledge graph embeddings based on numerical bias metrics. |
| Outcome: | The proposed framework can be extended to further bias definitions and applications. |
Knowledge Router: Learning Disentangled Representations for Knowledge Graphs (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing approaches to learning from relational patterns and structural information ignore the intrinsic complexity of KGs. |
| Approach: | They propose to learn latent properties of KG entities by using a neighborhood mechanism to disentangle the inner properties of each entity. |
| Outcome: | The proposed method significantly improves performance on key metrics on several benchmark datasets. |
A Simple LLM Framework for Long-Range Video Question-Answering (2024.emnlp-main)
Copied to clipboard
| Challenge: | a recent study has shown that short video understanding is not trivial due to the need for long-range temporal reasoning capabilities. |
| Approach: | They propose a language-based short- and long-range question-answering framework LLoVi . they propose 'multi-round summarization prompt' that asks the LLM to summarize the captions . |
| Outcome: | The proposed framework outperforms the state-of-the-art on the EgoSchema dataset and to grounded VideoQA. |
CogniVal in Action: An Interface for Customizable Cognitive Word Embedding Evaluation (2020.coling-demos)
Copied to clipboard
| Challenge: | Existing tools for evaluation of word embeddings are extrinsic and intrinsic methods, but they do not accurately reflect the meaning of words. |
| Approach: | They present a command-line interface for CogniVal with multiple improvements over the original framework and the possibility to evaluate custom embeddings against custom cognitive data sources. |
| Outcome: | The proposed system improves and extends the CogniVal framework and provides scalable and customized experiments. |
InstructPart: Task-Oriented Part Segmentation with Instruction Reasoning (2025.acl-long)
Copied to clipboard
Zifu Wan, Yaqi Xie, Ce Zhang, Zhiqiu Lin, Zihan Wang, Simon Stepputtis, Deva Ramanan, Katia P. Sycara
| Challenge: | Large multimodal foundation models perceive objects as indivisible, overlooking the components that constitute them. |
| Approach: | They propose a novel benchmark for large multimodal foundation models comprising hand-labeled part segmentation annotations and task-oriented instructions to evaluate their performance. |
| Outcome: | The proposed benchmark improves performance of current models in understanding and executing part-level tasks within everyday contexts. |
Control, Generate, Augment: A Scalable Framework for Multi-Attribute Text Generation (2020.findings-emnlp)
Copied to clipboard
| Challenge: | a new architecture for controlling, generating and augmenting text is being developed for supervised NLP tasks. |
| Approach: | They propose a conditional VAE architecture to control, generate, and augment text. |
| Outcome: | The proposed model shows high quality, diversity and attribute control in an ablation task. |