Papers by Qiusi Zhan

8 papers
GLEN: General-Purpose Event Detection for Thousands of Types (2023.emnlp-main)

Copied to clipboard

Challenge: ACE 2005 2 is the first large-scale event extraction dataset with 205K event mentions and 3,465 different types.
Approach: They propose to use the DWD Overlay to map PropBank rolesets to a large distantlysupervised training dataset with partial labels to make event extraction more accessible.
Outcome: The proposed model performs better than baselines including InstructGPT and ACE 2005 2 despite being 18 years old . key limitations of ACE include its small event ontology of 33 types, small dataset size of around 600 documents and restricted domain (with a significant portion concentrated on military conflicts).
MM-PoisonRAG: Disrupting Multimodal RAG with Local and Global Knowledge Poisoning Attacks (2026.acl-long)

Copied to clipboard

Challenge: Existing research exposes multimodal large language models to knowledge poisoning attacks . localized poisoning attack achieves up to 56% success rate even under restricted access . globalized poison attack completely disrupts model generation to 0% accuracy with just one poisoned content.
Approach: They propose a framework to study the vulnerability of multimodal RAG under knowledge poisoning attacks.
Outcome: The proposed framework exploits two new attack strategies on multimodal RAGs under knowledge poisoning.
SafeSearch: Do Not Trade Safety for Utility in LLM Search Agents (2026.findings-eacl)

Copied to clipboard

Challenge: Large language model (LLM) based search agents are more likely to produce harmful outputs than base models.
Approach: They propose a query-level shaping term that rewards safe queries and penalizes unsafe ones.
Outcome: The proposed approach reduces harmfulness by over 70% across three red-teaming datasets while producing safe, helpful responses.
ConFiguRe: Exploring Discourse-level Chinese Figures of Speech (2022.coling-1)

Copied to clipboard

Challenge: Figures of speech often deviate from their literal meanings to express deeper semantic implications.
Approach: They propose a concept of figurative unit, which is the carrier of a figure, and build a Chinese corpus for Contextualized Figure Recognition.
Outcome: The proposed model is based on 12 types of figures commonly used in Chinese . it shows that the proposed tasks are challenging for existing models .
Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents (2025.findings-naacl)

Copied to clipboard

Challenge: Large Language Models (LLMs) agents use external tools to interact with environments, but they also introduce security risks, such as indirect prompt injection (IPI) attacks.
Approach: They evaluate eight different defenses and bypass all of them using adaptive attacks.
Outcome: The proposed defenses achieve an attack success rate of over 50% and are robust against adaptive attacks.
InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents (2024.findings-acl)

Copied to clipboard

Challenge: Recent work has embodied LLMs as agents, but external content introduces the risk of indirect prompt injection (IPI) attacks.
Approach: They propose a benchmark to assess the vulnerability of tool-integrated LLM agents to IPI attacks.
Outcome: The proposed benchmark assesses the vulnerability of tool-integrated LLM agents to IPI attacks.
EA2E: Improving Consistency with Event Awareness for Document-Level Argument Extraction (2022.findings-naacl)

Copied to clipboard

Challenge: Recent work on document-level event argument extraction models each individual event in isolation and therefore causes inconsistency among extracted arguments across events.
Approach: They propose an event-aware argument extraction model with augmented context to improve consistency . they hypothesize that participants tend to play consistent roles across multiple events in a document .
Outcome: The proposed model improves consistency and accuracy of arguments extracted from documents.
Removing RLHF Protections in GPT-4 via Fine-Tuning (2024.naacl-short)

Copied to clipboard

Challenge: Large language models (LLMs) have increased in their capabilities, which increases their potential for dual use.
Approach: They show that fine-tuning can remove RLHFprotections with as few as 340 examples and a 95% success rate.
Outcome: The proposed method removes RLHFprotections with as few as 340 examples and a 95% success rate on non-censored outputs.

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