Zhenting Wang
YOU?
Author Swipe
View article: Debiasing LLMs by Masking Unfairness-Driving Attention Heads
Debiasing LLMs by Masking Unfairness-Driving Attention Heads Open
Large language models (LLMs) increasingly mediate decisions in domains where unfair treatment of demographic groups is unacceptable. Existing work probes when biased outputs appear, but gives little insight into the mechanisms that generat…
View article: Meaningless Tokens, Meaningful Gains: How Activation Shifts Enhance LLM Reasoning
Meaningless Tokens, Meaningful Gains: How Activation Shifts Enhance LLM Reasoning Open
Motivated by the puzzling observation that inserting long sequences of meaningless tokens before the query prompt can consistently enhance LLM reasoning performance, this work analyzes the underlying mechanism driving this phenomenon and b…
View article: EDITOR: Effective and Interpretable Prompt Inversion for Text-to-Image Diffusion Models
EDITOR: Effective and Interpretable Prompt Inversion for Text-to-Image Diffusion Models Open
Text-to-image generation models~(e.g., Stable Diffusion) have achieved significant advancements, enabling the creation of high-quality and realistic images based on textual descriptions. Prompt inversion, the task of identifying the textua…
View article: DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-training
DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-training Open
Recent advances in reinforcement learning (RL)-based post-training have led to notable improvements in large language models (LLMs), particularly in enhancing their reasoning capabilities to handle complex tasks. However, most existing met…
View article: Can Large Vision-Language Models Detect Images Copyright Infringement from GenAI?
Can Large Vision-Language Models Detect Images Copyright Infringement from GenAI? Open
Generative AI models, renowned for their ability to synthesize high-quality content, have sparked growing concerns over the improper generation of copyright-protected material. While recent studies have proposed various approaches to addre…
View article: ADO: Automatic Data Optimization for Inputs in LLM Prompts
ADO: Automatic Data Optimization for Inputs in LLM Prompts Open
This study explores a novel approach to enhance the performance of Large Language Models (LLMs) through the optimization of input data within prompts. While previous research has primarily focused on refining instruction components and aug…
View article: Identify drug-drug interactions via deep learning: A real world study
Identify drug-drug interactions via deep learning: A real world study Open
Identifying drug-drug interactions (DDIs) is essential to prevent adverse effects from polypharmacy. Although deep learning has advanced DDI identification, the gap between powerful models and their lack of clinical application and evaluat…
View article: Token-Budget-Aware LLM Reasoning
Token-Budget-Aware LLM Reasoning Open
View article: EmojiPrompt: Generative Prompt Obfuscation for Privacy-Preserving Communication with Cloud-based LLMs
EmojiPrompt: Generative Prompt Obfuscation for Privacy-Preserving Communication with Cloud-based LLMs Open
View article: An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer
An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer Open
View article: Data-centric NLP Backdoor Defense from the Lens of Memorization
Data-centric NLP Backdoor Defense from the Lens of Memorization Open
View article: Auto-Prompt Generation is Not Robust: Prompt Optimization Driven by Pseudo Gradient
Auto-Prompt Generation is Not Robust: Prompt Optimization Driven by Pseudo Gradient Open
While automatic prompt generation methods have recently received significant attention, their robustness remains poorly understood. In this paper, we introduce PertBench, a comprehensive benchmark dataset that includes a wide range of inpu…
View article: Token-Budget-Aware LLM Reasoning
Token-Budget-Aware LLM Reasoning Open
Reasoning is critical for large language models (LLMs) to excel in a wide range of tasks. While methods like Chain-of-Thought (CoT) reasoning and enhance LLM performance by decomposing problems into intermediate steps, they also incur sign…
View article: The characteristics and analysis of the complete chloroplast genome of <i>Hemerocallis</i> cultivar Small orange lamp 2019 (Asphodelaceae)
The characteristics and analysis of the complete chloroplast genome of <i>Hemerocallis</i> cultivar Small orange lamp 2019 (Asphodelaceae) Open
Hemerocallis cultivar Small orange lamp is a hybrid variety. Its whole chloroplast genome was 156,053 bp in size, consisting of 135 genes in total, including 89 mRNA genes, 38 tRNA genes, and 8 rRNA genes. The chloroplast genome con…
View article: Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction
Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction Open
Prevailing Multimodal Large Language Models (MLLMs) encode the input image(s) as vision tokens and feed them into the language backbone, similar to how Large Language Models (LLMs) process the text tokens. However, the number of vision tok…
View article: Continuous Concepts Removal in Text-to-image Diffusion Models
Continuous Concepts Removal in Text-to-image Diffusion Models Open
Text-to-image diffusion models have shown an impressive ability to generate high-quality images from input textual descriptions. However, concerns have been raised about the potential for these models to create content that infringes on co…
View article: Desertification Mitigation in Northern China Was Promoted by Climate Drivers after 2000
Desertification Mitigation in Northern China Was Promoted by Climate Drivers after 2000 Open
Desertification greatly threatens the ecological environment and sustainable development over approximately 30% of global land. In this study, the contributions of climate drivers and human activity in shaping the desertification process f…
View article: Data-centric NLP Backdoor Defense from the Lens of Memorization
Data-centric NLP Backdoor Defense from the Lens of Memorization Open
Backdoor attack is a severe threat to the trustworthiness of DNN-based language models. In this paper, we first extend the definition of memorization of language models from sample-wise to more fine-grained sentence element-wise (e.g., wor…
View article: An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer
An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer Open
Despite prior safety alignment efforts, mainstream LLMs can still generate harmful and unethical content when subjected to jailbreaking attacks. Existing jailbreaking methods fall into two main categories: template-based and optimization-b…
View article: Visual Agents as Fast and Slow Thinkers
Visual Agents as Fast and Slow Thinkers Open
Achieving human-level intelligence requires refining cognitive distinctions between System 1 and System 2 thinking. While contemporary AI, driven by large language models, demonstrates human-like traits, it falls short of genuine cognition…
View article: When AI Meets Finance (StockAgent): Large Language Model-based Stock Trading in Simulated Real-world Environments
When AI Meets Finance (StockAgent): Large Language Model-based Stock Trading in Simulated Real-world Environments Open
Can AI Agents simulate real-world trading environments to investigate the impact of external factors on stock trading activities (e.g., macroeconomics, policy changes, company fundamentals, and global events)? These factors, which frequent…
View article: APEER: Automatic Prompt Engineering Enhances Large Language Model Reranking
APEER: Automatic Prompt Engineering Enhances Large Language Model Reranking Open
Large Language Models (LLMs) have significantly enhanced Information Retrieval (IR) across various modules, such as reranking. Despite impressive performance, current zero-shot relevance ranking with LLMs heavily relies on human prompt eng…
View article: Evaluating and Mitigating IP Infringement in Visual Generative AI
Evaluating and Mitigating IP Infringement in Visual Generative AI Open
The popularity of visual generative AI models like DALL-E 3, Stable Diffusion XL, Stable Video Diffusion, and Sora has been increasing. Through extensive evaluation, we discovered that the state-of-the-art visual generative models can gene…
View article: How to Trace Latent Generative Model Generated Images without Artificial Watermark?
How to Trace Latent Generative Model Generated Images without Artificial Watermark? Open
Latent generative models (e.g., Stable Diffusion) have become more and more popular, but concerns have arisen regarding potential misuse related to images generated by these models. It is, therefore, necessary to analyze the origin of imag…
View article: Some statistical properties of aeolian saltation
Some statistical properties of aeolian saltation Open
Aeolian sediment transport is a process that commonly occurs on celestial bodies with atmospheric layers and solid surfaces. At present, it is very difficult to predict the instantaneous mass flux accurately. For the purpose of statistical…
View article: Alteration-free and Model-agnostic Origin Attribution of Generated Images
Alteration-free and Model-agnostic Origin Attribution of Generated Images Open
Recently, there has been a growing attention in image generation models. However, concerns have emerged regarding potential misuse and intellectual property (IP) infringement associated with these models. Therefore, it is necessary to anal…
View article: NOTABLE: Transferable Backdoor Attacks Against Prompt-based NLP Models
NOTABLE: Transferable Backdoor Attacks Against Prompt-based NLP Models Open
Prompt-based learning is vulnerable to backdoor attacks. Existing backdoor attacks against prompt-based models consider injecting backdoors into the entire embedding layers or word embedding vectors. Such attacks can be easily affected by …
View article: UNICORN: A Unified Backdoor Trigger Inversion Framework
UNICORN: A Unified Backdoor Trigger Inversion Framework Open
The backdoor attack, where the adversary uses inputs stamped with triggers (e.g., a patch) to activate pre-planted malicious behaviors, is a severe threat to Deep Neural Network (DNN) models. Trigger inversion is an effective way of identi…
View article: Unintended consequences of combating desertification in China
Unintended consequences of combating desertification in China Open
Since the early 2000s, China has carried out extensive “grain-for-green” and grazing exclusion practices to combat desertification in the desertification-prone region (DPR). However, the environmental and socioeconomic impacts of these pra…
View article: A Novel Methylation Marker NRN1 plus TERT and FGFR3 Mutation Using Urine Sediment Enables the Detection of Urothelial Bladder Carcinoma
A Novel Methylation Marker NRN1 plus TERT and FGFR3 Mutation Using Urine Sediment Enables the Detection of Urothelial Bladder Carcinoma Open
Background: Aberrant DNA methylation is an early event during tumorigenesis. In the present study, we aimed to construct a methylation diagnostic tool using urine sediment for the detection of urothelial bladder carcinoma, and improved the…