Papers by Haohan Wang
Enhancing Extractive Question Answering in Multiparty Dialogues with Logical Inference Memory Network (2025.coling-main)
Copied to clipboard
| Challenge: | Existing models for multiparty dialogue question answering (QA) do not consider logical inference relations in multiparty dialogs, leading to suboptimal performance. |
| Approach: | They propose a memory network with logical inference for extractive QA in multiparty dialogues. |
| Outcome: | The proposed model achieves state-of-the-art on Molweni and FriendsQA benchmarks. |
Measure and Improve Robustness in NLP Models: A Survey (2022.naacl-main)
Copied to clipboard
| Challenge: | Despite the performance gains, NLP models are still fragile and brittle to out-of-domain data, adversarial attacks, or small perturbation to the input. |
| Approach: | They propose a survey of how to define, measure and improve robustness in NLP by connecting multiple definitions of robustness and identifying failures. |
| Outcome: | The proposed models are robust against unseen or challenging scenarios, but are still fragile and brittle to out-of-domain data and adversarial attacks. |
MemeCLIP: Leveraging CLIP Representations for Multimodal Meme Classification (2024.emnlp-main)
Copied to clipboard
| Challenge: | a novel dataset of text-embedded images associated with the LGBTQ+ Pride movement is presented in this paper . a new framework for analyzing text-based images is proposed to address this challenge . |
| Approach: | They propose a new dataset for machine learning that includes hate, targets of hate, stance, humor and a framework for efficient downstream learning while preserving the knowledge of the pre-trained CLIP model. |
| Outcome: | The proposed framework achieves superior performance on two real-world datasets. |
Crafting Adversarial Inputs for Large Vision-Language Models Using Black-Box Optimization (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing white-box jailbreak methods require full model accessibility and require computational costs. |
| Approach: | They propose a black-box jailbreak attack using Zeroth-Order optimization using ZO-SPSA. |
| Outcome: | The proposed method achieves highest jailbreak success rate on three LVLMs, including InstructBLIP, LLaVA and MiniGPT-4. |
Exploring the Vulnerability of the Content Moderation Guardrail in Large Language Models via Intent Manipulation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Prior work has shown that intent detection enhances LLMs’ moderation guardrails, but the robustness of these guardrail mechanisms under malicious manipulations remains under-explored. |
| Approach: | They propose a two-stage intent-based prompt-refinement framework that first transforms harmful inquiries into structured outlines and further reframes them into declarative-style narratives. |
| Outcome: | The proposed framework outperforms several cutting-edge jailbreak methods and evades even advanced Intent Analysis (IA) and Chain-of-Thought (CoT)-based defenses. |
Learning to Conceal Risk: Controllable Multi-turn Red Teaming for LLMs in the Financial Domain (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly deployed in high-stakes domains such as finance where unsafe behavior can lead to serious regulatory risks. |
| Approach: | They propose a black-box multi-turn risk-concealed redteaming framework that progressively conceals surface-level risk while exploiting regulatory-violating behaviors. |
| Outcome: | Experiments on nine widely used LLMs show that the proposed framework achieves 93.19% average attack success rate (ASR) and improves the average ASR to 95.00%. |
Unlearn Dataset Bias in Natural Language Inference by Fitting the Residual (D19-61)
Copied to clipboard
| Challenge: | Statistical natural language inference models are susceptible to learning dataset bias. |
| Approach: | They propose a debiasing algorithm that debiases models that use only known dataset biases . they use two benchmark datasets to train three high-performing NLI models . |
| Outcome: | The proposed learning objective improves model performance on challenge datasets while maintaining reasonable performance on original datasets. |
Now You Hear Me: Audio Narrative Attacks Against Large Audio–Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing jailbreaks against large audio-language models fall into two categories . early work converted text-based prompts into synthetic speech, while subsequent work introduced minor acoustic variations such as accent shifts, phonetic spellings, or stress patterns. |
| Approach: | They propose a text-to-audio jailbreak that embeds disallowed directives within a narrative-style audio stream. |
| Outcome: | The proposed attack exploits structural and acoustic properties of a text-to-audio model . it achieves 98.26% success rate, significantly exceeding baselines for text-based models . |