Papers by Jiahui Geng

17 papers
\mathsf{Con Instruction}: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities (2025.acl-long)

Copied to clipboard

Challenge: Existing attacks communicate instruction through text, accompanied by a toxic image or audio . a novel gray-box attack method generates adversarial images or audio to convey harmful instructions to MLLMs .
Approach: They propose a gray-box attack method that generates adversarial images or audio to convey specific harmful instructions to MLLMs by following non-textual instruction.
Outcome: The proposed method achieves highest success rates on visual and audio-language models . larger models are more susceptible toCon Instruction, compared to their underlying models - the results will be released .
CoQuIR: A Comprehensive Benchmark for Code Quality-Aware Information Retrieval (2026.acl-long)

Copied to clipboard

Challenge: Existing benchmarks focus on functional relevance while neglecting code quality.
Approach: They propose a multilingual benchmark to evaluate quality-aware code retrieval . they include fine-grained quality annotations over 42,725 queries and 134,907 code snippets .
Outcome: The proposed benchmarks show that state-of-the-art models fail to separate buggy or insecure code from robust counterparts.
SGPVT: Self-Generated Proximal Visual Tokens for Mitigating Proximal Collateral Damage in MLLM Unlearning (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches focus on general utility metrics, overlooking the preservation of semantically related concepts.
Approach: They propose a method that introduces self-generated proximal visual tokens to prevent forgetting vulnerability.
Outcome: The proposed framework outperforms existing methods in preserving semantically related concepts while achieving effective target unlearning.
A Survey of Confidence Estimation and Calibration in Large Language Models (2024.naacl-long)

Copied to clipboard

Challenge: Large language models (LLMs) have demonstrated impressive capabilities across a wide range of tasks in various domains, but they can be unreliable due to factual errors in their generations.
Approach: They summarize recent advances in LLM confidence estimation and calibration and outline their main lessons learned.
Outcome: The proposed methods can be used to assess the reliability of models and to calibrate them across tasks.
OpenFactCheck: Building, Benchmarking Customized Fact-Checking Systems and Evaluating the Factuality of Claims and LLMs (2025.coling-main)

Copied to clipboard

Challenge: Large language models (LLMs) generate naturallysounding answers over a broad range of human inquiries, but they still produce content that deviates from real-world facts.
Approach: They propose a framework for building customized automatic fact-checking systems, benchmarking their accuracy, evaluating factuality of LLMs, and verifying claims in a document.
Outcome: The proposed framework assesses the factuality of free-form responses in open domains and evaluates factually of LLMs.
Marco-Bench-MIF: On Multilingual Instruction-Following Capability of Large Language (2025.acl-long)

Copied to clipboard

Challenge: Existing datasets for instruction-following are monolingual and centered on English . existing data are unable to capture linguistic and cultural subtle differences .
Approach: They propose an extension of IFEval to a localized multilingual version called Marco-Bench-MIF . their benchmark addresses linguistic constraints and cultural references via translation and verification .
Outcome: The proposed extension of IFEval to a localized multilingual version covers 30 languages with varying levels of localization.
Is Human-Like Text Liked by Humans? Multilingual Human Detection and Preference Against AI (2026.acl-long)

Copied to clipboard

Challenge: Prior studies have shown that distinguishing text generated by Large Language Models from human-written text is challenging for humans and often no better than random guessing.
Approach: They conduct extensive case study to determine the upper bound of human detection accuracy.
Outcome: The findings challenge previous conclusions on human detection accuracy across languages and domains.
LLM-DetectAIve: a Tool for Fine-Grained Machine-Generated Text Detection (2024.emnlp-demo)

Copied to clipboard

Challenge: a large number of machine-generated texts are often hard to distinguish between human-written and machine-generated text . this raises concerns about potential misuse, especially within educational and academic domains .
Approach: They propose a system that can detect whether a text is human-written or machine-generated . they use a fine-grained classification schema to identify the use of machine-generated text .
Outcome: The proposed system can distinguish between human-written and machine-generated text . it can detect attempts to obfuscate the fact that a text was machine- generated .
Factcheck-Bench: Fine-Grained Evaluation Benchmark for Automatic Fact-checkers (2024.findings-emnlp)

Copied to clipboard

Challenge: Large language models generate naturally sounding answers over a broad range of human inquiries, but they often generate answers that contradict real-world facts.
Approach: They propose a framework for annotating and evaluating the factuality of large language models . they propose 'factcheck-bench' which provides a multi-stage annotation scheme .
Outcome: The proposed framework outperforms several popular LLM fact-checkers in claim, sentence, and document levels.
FIRE: Fact-checking with Iterative Retrieval and Verification (2025.findings-naacl)

Copied to clipboard

Challenge: Fact-checking long-form text is challenging, and breaking it down into multiple atomic claims is not cost-effective.
Approach: They propose a novel agent-based framework that integrates evidence retrieval and claim verification in an iterative manner.
Outcome: The proposed framework reduces large language model (LLM) costs by an average of 7.6 times and search costs by 16.5 times while retaining the same performance.
OpenFactCheck: A Unified Framework for Factuality Evaluation of LLMs (2024.emnlp-demo)

Copied to clipboard

Challenge: Large language models (LLMs) often produce content that deviates from real-world facts.
Approach: They developed a unified framework to assess the factuality of large language models . open-sourced framework is publicly available as a Python library and web service .
Outcome: OpenFactCheck is open-sourced and publicly released as a Python library and web service.
Improving Unsupervised Word-by-Word Translation with Language Model and Denoising Autoencoder (D18-1)

Copied to clipboard

Challenge: Unsupervised learning of cross-lingual word embeddings has fundamental limitations in translating sentences.
Approach: They propose a method to improve word-by-word translation of cross-lingual embeddings using monolingual corpora without any back-translation.
Outcome: The proposed system surpasses state-of-the-art unsupervised translation systems without costly iterative training.
VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration (2025.findings-acl)

Copied to clipboard

Challenge: Existing safety calibration methods focus on model undersafety, where the model responds to hazardous queries, while neglecting oversafetiness, where models refuse to answer safe queries.
Approach: They propose safety calibration which addresses both undersafety and oversafetiness by comparing model responses to a novel dataset of 3,600 image-text pairs.
Outcome: The proposed methods have been used to evaluate safety calibration across image-centric and text-centric scenarios.
Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models (2025.acl-long)

Copied to clipboard

Challenge: Understanding how jailbreaking works remains limited, hindering the development of effective defense strategies.
Approach: They propose a new mechanism that adaptively constrains activations within the safety boundary and propose 'Activation Boundary Defense' to enhance its effectiveness.
Outcome: The proposed defense achieves an average Defense Success Rate (DSR) of over 98% against various jailbreak attacks, with less than 2% impact on the model’s general capabilities.
HD-NDEs: Neural Differential Equations for Hallucination Detection in LLMs (2025.acl-long)

Copied to clipboard

Challenge: Hallucination is a significant challenge for large language models, but current methods struggle when non-factual information arises in the early or mid-sequence of outputs, reducing their reliability.
Approach: They propose a method that captures the full dynamics of large language models by using neural differential equations to assess the truthfulness of statements.
Outcome: The proposed method achieves 14% improvement in AUC-ROC on the True-False dataset compared to state-of-the-art methods.
Reference-free Hallucination Detection for Large Vision-Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Large vision-language models exhibit excellent ability in language understanding, question answering, and conversations of visual inputs, but they are prone to producing hallucinations.
Approach: They propose to use supervised uncertainty quantification methods to detect hallucinations in large vision-language models.
Outcome: The proposed methods outperform the others in detecting hallucinations on four representative LVLMs across two different tasks.

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