Papers by Zihan Wu

18 papers
IMOL: Incomplete-Modality-Tolerant Learning for Multi-Domain Fake News Video Detection (2025.acl-long)

Copied to clipboard

Challenge: Existing methods for fake news video detection focus on a specific domain and assume multiple modalities.
Approach: They propose an incomplete-modality-tolerant learning framework for fake news video detection . they use cross-modal consistency to reconstruct missing modalities and transferable knowledge through cross-sample reasoning .
Outcome: The proposed framework improves performance and robustness of multi-domain fake news video detection while generalizing to unseen domains under incomplete modality conditions.
From Detection to Understanding: Multi-Turn Reasoning for Video Misinformation Analysis (2026.acl-long)

Copied to clipboard

Challenge: Existing benchmarks focus on binary veracity judgments and do not evaluate process-level justifications for misinformation models.
Approach: They propose a video misinformation analysis benchmark that assesses reasoning in video misinterpretation.
Outcome: The proposed framework improves reasoning accuracy and explanation quality compared to existing models . it covers 12 fine-grained deception categories and progresses from perceptual attribution to intent and persuasion analysis.
Image Difference Captioning via Adversarial Preference Optimization (2025.emnlp-main)

Copied to clipboard

Challenge: Existing supervised approaches to image difference captioning overfit to dataset-specific language patterns and fail to capture accurate preferences.
Approach: They propose an adversarial direct preference optimization framework that aligns captioning policy with pairwise difference preferences via Direct Preference Optimization.
Outcome: The proposed approach outperforms baselines on benchmark IDC datasets in generating fine-grained and accurate difference descriptions.
LegalDiscourse: Interpreting When Laws Apply and To Whom (2024.naacl-long)

Copied to clipboard

Challenge: Recent advances in NLP and information retrieval have already enabled practical applications.
Approach: They propose a 'discourse' taxonomy for span-and-relation parsing of legal texts . they use a dataset of 602 state-level law paragraphs with 3,715 discourse spans and 1,671 relations to investigate the increase in liquor licenses and decrease in applicable laws.
Outcome: The proposed model performs poorly at span identification and relation classification, but lags far below human level.
HyperMoE: Towards Better Mixture of Experts via Transferring Among Experts (2024.acl-long)

Copied to clipboard

Challenge: Existing methods for enhancing performance through increased use of expert knowledge often result in diminishing sparsity during expert selection.
Approach: They propose a framework that integrates the computational processes of MoE with the concept of knowledge transferring in multi-task learning.
Outcome: The proposed framework outperforms existing methods under identical conditions concerning the number of experts.
Benchmarking Foundation Models with Retrieval-Augmented Generation in Olympic-Level Physics Problem Solving (2025.findings-emnlp)

Copied to clipboard

Challenge: a new study examines the potential of retrieval-augmented generation (RAG) with foundation models to enhance expert-level reasoning.
Approach: They introduce PhoPile, a high-quality multimodal dataset specifically designed for Olympiad-level physics.
Outcome: The proposed model can be used to solve Olympiad-level physics problems.
Classic4Children: Adapting Chinese Literary Classics for Children with Large Language Model (2025.findings-naacl)

Copied to clipboard

Challenge: Recent large language models (LLMs) overlook children’s reading preferences, which poses challenges in CLA.
Approach: They propose a method that augments large language models with children's reading preferences for adaptation by obtaining characters' personalities and narrative structure as additional information for fine-grained instruction tuning.
Outcome: The proposed method significantly improves performance in automatic and human evaluation.
Pre-training Language Model as a Multi-perspective Course Learner (2023.findings-acl)

Copied to clipboard

Challenge: Experimental results show that our method significantly improves ELECTRA’s average performance by 2.8% and 3.2% absolute points respectively on GLUE and SQuAD 2.0 benchmarks.
Approach: They propose a multi-perspective course learning method to fetch many degrees and visual angles for sample-efficient pre-training and to fully leverage the relationship between generator and discriminator.
Outcome: The proposed method improves ELECTRA's performance on GLUE and SQuAD 2.0 benchmarks and overshadows recent advanced ELECL-style models under the same settings.
DualGuard: A Parameter Space Transformation Approach for Bidirectional Defense in Split-Based LLM Fine-Tuning (2025.acl-long)

Copied to clipboard

Challenge: Existing defense methods for large language model fine-tuning (LLM-FT) sacrifice task-specific performance under privacy constraints.
Approach: They propose a bidirectional defense mechanism that uses a local warm-up parameter transformation to alter client-side model parameters before training.
Outcome: The proposed defense mechanism outperforms current defense methods while maintaining task performance.
Sinkhorn Distance Minimization for Knowledge Distillation (2024.lrec-main)

Copied to clipboard

Challenge: Existing knowledge distillation methods investigate divergence measures but fail to deliver effective supervision when few distribution overlap exists between teacher and student.
Approach: They propose a knowledge distillation method that exploits the Sinkhorn distance to ensure a nuanced assessment of the disparity between teacher and student distributions.
Outcome: The proposed method outperforms state-of-the-art methods on all kinds of LLMs with encoder-only, encoder decoder, and decoded architectures.
Multi-step Problem Solving Through a Verifier: An Empirical Analysis on Model-induced Process Supervision (2024.findings-emnlp)

Copied to clipboard

Challenge: a method for process supervision has shown significant improvements in multi-step problem solving . despite the advances in process supervision, there are still easily observable mistakes in state-of-the-art LLMs.
Approach: They propose a method for automating data curation by using a trained verifier to evaluate intermediate steps generated by a reasoner.
Outcome: The proposed method improves the performance of PaLM 2 on math and coding tasks.
Enhancing Learning-Based Binary Code Similarity Detection Model through Adversarial Training with Multiple Function Variants (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing Learning-Based Binary Code Similarity Detection (LB-BCSD) methods exhibit lower accuracy in recognizing functions with the same functionality but different implementations.
Approach: They propose a gradient-guided adversarial attack method based on critical code called FuncFooler which perturbs critical code to generate multiple variants of the same function.
Outcome: The proposed method increases the accuracy of the current Learning-Based Binary Code Similarity Detection (LB-BCSD) model by 5%-7%.
Enhancing the General Agent Capabilities of Low-Paramter LLMs through Tuning and Multi-Branch Reasoning (2024.findings-naacl)

Copied to clipboard

Challenge: Open-source pre-trained Large Language Models exhibit strong language understanding and generation capabilities, making them highly successful in a variety of tasks.
Approach: They propose a method to construct agent-specific data using GPT-4 and supervised fine-tuning . they find that supervised tunning can significantly reduce hallucination outputs and formatting errors in agent tasks .
Outcome: The proposed method improves on five agent tasks of AgentBench.
ALLabel: Three-stage Active Learning for LLM-based Entity Recognition using Demonstration Retrieval (2025.emnlp-main)

Copied to clipboard

Challenge: Large language models (LLMs) are increasingly used to solve the entity recognition task.
Approach: They propose a framework to select the most informative and representative samples for LLM in-context learning.
Outcome: The proposed framework outperforms baselines on three specialized domain datasets.
Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Existing studies on parameter-efficient fine-tuning (PEFT) for dense-architecture LLMs are lacking.
Approach: They propose an expert-specialized fine-tuning method that tunes the experts most relevant to downstream tasks while freezing the other experts.
Outcome: The proposed method matches or surpasses full-parameter fine-tuning.
No Free Lunch: Retrieval-Augmented Generation Undermines Fairness in LLMs, Even for Vigilant Users (2025.findings-emnlp)

Copied to clipboard

Challenge: Retrieval-augmented generation is widely adopted for its effectiveness and cost-efficiency in mitigating hallucinations.
Approach: They propose a practical three-level threat model from the perspective of user fairness awareness.
Outcome: The proposed model shows that RAG can undermine fairness alignment without fine-tuning or retraining.
MulVul: Retrieval-augmented Multi-Agent Code Vulnerability Detection via Cross-Model Prompt Evolution (2026.acl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) struggle to automate real-world vulnerability detection due to the heterogeneity of vulnerability patterns and manual prompt engineering for massive weakness categories is unscalable.
Approach: They propose a retrieval-augmented multi-agent framework for precise and broad-coverage vulnerability detection using a coarse-to-fine strategy.
Outcome: The proposed framework outperforms the baseline model on 130 CWE types and achieves 34.79% Macro-F1 performance.
CuBridge: An LLM-Based Framework for Understanding and Reconstructing High-Performance Attention Kernels (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to support diverse attention variants trade performance for flexibility . expert-written kernels achieve high efficiency but are difficult to adapt .
Approach: They propose a framework that adapts expert-written attention kernels to GPUs . they use a structured lift–transfer–lower workflow to make execution explicit .
Outcome: The proposed framework outperforms existing frameworks and compilers on diverse variants and GPU platforms.

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