Mi Zhang

Improving the Robustness of Wasserstein Embedding by Adversarial PAC-Bayesian Learning

2020 · Proceedings of the AAAI Conference on Artificial Intelligence

Node embedding is a crucial task in graph analysis. Recently, several methods are proposed to embed a node as a distribution rather than a vector to capture more information. Although these methods achieved noticeable improvements, …

Convolution over Hierarchical Syntactic and Lexical Graphs for Aspect Level Sentiment Analysis

2020

The state-of-the-art methods in aspect-level sentiment classification have leveraged the graph based models to incorporate the syntactic structure of a sentence. While being effective, these methods ignore the corpus level word co-occurrence information, which reflect …

CATE: Computation-aware Neural Architecture Encoding with Transformers

2021 · arXiv (Cornell University)

Recent works (White et al., 2020a; Yan et al., 2020) demonstrate the importance of architecture encodings in Neural Architecture Search (NAS). These encodings encode either structure or computation information of the neural architectures. Compared to …

Differential Cryptanalysis of TweGIFT-128 Based on Neural Network

2021

It is a new trend of cryptographic analysis to realize automatic analysis on cryptographic algorithms by means of deep learning in recent years. TweGIFT-128 algorithm is an instantiation tweak block cipher algorithm for encryption authentication …

MaSS: Model-agnostic, Semantic and Stealthy Data Poisoning Attack on Knowledge Graph Embedding

2023

Open-source knowledge graphs are attracting increasing attention. Nevertheless, the openness also raises the concern of data poisoning attacks, that is, the attacker could submit malicious facts to bias the prediction of knowledge graph embedding (KGE) …

Enhancing Test-Time Scaling of Large Language Models with Hierarchical Retrieval-Augmented MCTS

2025 · arXiv (Cornell University)

Test-time scaling has emerged as a promising paradigm in language modeling, leveraging additional computational resources at inference time to enhance model performance. In this work, we introduce R2-LLMs, a novel and versatile hierarchical retrieval-augmented reasoning …

SmartSight: Mitigating Hallucination in Video-LLMs Without Compromising Video Understanding via Temporal Attention Collapse

2026 · Proceedings of the AAAI Conference on Artificial Intelligence

Despite Video Large Language Models (Video-LLMs) having rapidly advanced in recent years, perceptual hallucinations pose a substantial safety risk, which severely restricts their real-world applicability. While several methods for hallucination mitigation have been proposed, they …