Papers by Zeyu Huang

14 papers
PaddleSpeech: An Easy-to-Use All-in-One Speech Toolkit (2022.naacl-demo)

Copied to clipboard

Challenge: PaddleSpeech is an open-source speech toolkit that supports speech-to-text and text-to speech tasks.
Approach: They describe the design philosophy and core architecture of PaddleSpeech to support several essential speech-to-text and text-to speech tasks.
Outcome: The proposed framework achieves competitive or state-of-the-art performance on various speech datasets and implements the most popular methods.
A Regularization Approach for Incorporating Event Knowledge and Coreference Relations into Neural Discourse Parsing (D19-1)

Copied to clipboard

Challenge: Existing approaches to discourse parsing use commonsense knowledge and linguistic constraints to integrate them into neural network models.
Approach: They propose a knowledge regularization approach that integrates linguistic constraints with contexts for deriving word representations.
Outcome: The proposed approach outperforms previous systems on the benchmark dataset PDTB for discourse parsing.
Weakly Supervised Subevent Knowledge Acquisition (2020.emnlp-main)

Copied to clipboard

Challenge: Subevents elaborate an event and exist in event descriptions.
Approach: They propose a weakly supervised approach to extract subevent relation tuples from text . they then use the initial seed subeven pairs to train a contextual classifier .
Outcome: The proposed method is high quality and covers a wide range of event types.
Demons in the Detail: On Implementing Load Balancing Loss for Training Specialized Mixture-of-Expert Models (2025.acl-long)

Copied to clipboard

Challenge: Existing Mixture-of-Experts training frameworks use a micro-batch to calculate LBL . micro-batches are restricted to a single sequence, preventing expert specialization .
Approach: They propose to use a global-batch to loosen the load balance constraint for MoEs models . they propose to synchronize fi across micro-batches and then use it to calculate the LBL .
Outcome: The proposed global-batch LBL improves the domain specialization of experts . the micro-battery LBL is almost at the sequence level, and the router is pushed to distribute the token evenly .
VaseVQA: Multimodal Agent and Benchmark for Ancient Greek Pottery (2026.findings-eacl)

Copied to clipboard

Challenge: MLLMs that use domain-specific data are limited in understanding cultural heritage artifacts such as ancient Greek pottery . supervised fine-tuning improves adaptation to domain knowledge, but it struggles with deeper reasoning tasks.
Approach: They propose a visual question-answer tool that augments SFT with reinforcement learning using verifiable rewards.
Outcome: The proposed model outperforms baseline models on reasoning-intensive questions on ancient Greek pottery.
Unlocking Emergent Modularity in Large Language Models (2024.naacl-long)

Copied to clipboard

Challenge: Existing MNNs are explicit, with predefined modular architectures and individual modules expected to implement distinct functions.
Approach: They propose to unlock emergent modularity in language models by fine-tuning them as Mixture-of-Experts (MoEs) EMoE is robust to various configurations and can scale up to Large Language Models .
Outcome: The proposed models can be fine-tuned as Mixture-of-Expert (MoE) counterparts without introducing any extra parameters.
A Closer Look into Mixture-of-Experts in Large Language Models (2025.findings-naacl)

Copied to clipboard

Challenge: Mixture-of-experts (MoE) architectures are gaining increasing attention for their unique properties and remarkable performance.
Approach: They propose a mixture-of-experts architecture that allows for model scaling without sacrificing computational efficiency.
Outcome: The proposed model increases model size without sacrificing computational efficiency . the proposed model is modular and can be used by a broad spectrum of practitioners .
A Joint Model for Structure-based News Genre Classification with Application to Text Summarization (2021.findings-acl)

Copied to clipboard

Challenge: Existing models for structure-based news genre classification identify news structure types and news elements . authors show that the model outperforms variants that perform two tasks independently .
Approach: They propose a joint model that identifies one of four commonly used news structures for a news article and recognizes a sequence of news elements within the article that define the corresponding news structure.
Outcome: The proposed model outperforms variants that perform two tasks independently . it predicts news structure type and news elements and improves text summarization .
Unlocking Continual Learning Abilities in Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to learning models (LMs) incorporate old task data or task-wise inductive bias into LMs, but old data and accurate task information are often unavailable or costly to collect.
Approach: They propose a rehearsal-free method that updates model parameters with large magnitudes . they found that the L1-normalized magnitude distribution is different when different task data is used .
Outcome: The proposed method improves accuracy and performance on four CL benchmarks.
Improving Implicit Discourse Relation Classification by Modeling Inter-dependencies of Discourse Units in a Paragraph (N18-1)

Copied to clipboard

Challenge: Existing methods for predicting implicit discourse relations ignore wider paragraph contexts beyond the two discourse units examined for a discourse relation prediction.
Approach: They propose a paragraph-level neural network that models inter-dependencies between discourse units and discourse relation continuity and patterns and predicts a sequence of discourse relations in a sentence.
Outcome: The proposed model outperforms state-of-the-art systems on the benchmark corpus of PDTB.
Building Context-aware Clause Representations for Situation Entity Type Classification (D18-1)

Copied to clipboard

Challenge: Existing models for categorizing clauses based on situation entity types do not provide accurate results.
Approach: They propose to build context-aware clause representations for predicting situation entity types of clauses by modeling context influences and inter-dependencies of clause.
Outcome: The proposed model achieves state-of-the-art performance on genre-rich dataset MASC+Wiki .
DPGA-TextSyn: Differentially Private Genetic Algorithm for Synthetic Text Generation (2025.findings-acl)

Copied to clipboard

Challenge: Existing methods to fine-tune large language models pose privacy risks . researchers have synthesized data with strong generation capabilities closed-source LLMs to alleviate this problem .
Approach: They propose to combine general LLMs with genetic algorithm to produce relevant and diverse synthetic text under differential privacy constraints.
Outcome: The proposed method significantly improves the performance of the model in downstream tasks while ensuring privacy.
Mixture of Attention Heads: Selecting Attention Heads Per Token (2022.emnlp-main)

Copied to clipboard

Challenge: Mixture-of-Experts (MoE) networks have been proposed as an efficient way to scale up model capacity and implement conditional computing.
Approach: They propose a new architecture that combines multi-head attention with the MoE mechanism and a sparsely gated architecture that allows for faster computations.
Outcome: The proposed architecture can scale up the number of attention heads and the number parameters while preserving computational efficiency.
TRACE: Evaluating Execution Efficiency of LLM-Based Code Translation (2026.acl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have improved the functional correctness of code translation, but execution efficiency remains overlooked.
Approach: They propose a benchmark to explicitly assess execution efficiency in LLM-translated code.
Outcome: The proposed benchmark identifies that execution efficiency is an essential dimension of code translation . the results highlight that correctness and efficiency are often misaligned .

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations