Papers by Jiayi Fu

7 papers
SocialGaze: Improving the Integration of Human Social Norms in Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Increasingly, large language models (LLMs) are able to understand and rationalize socially acceptable behaviors, but they are often misaligned with human consensus.
Approach: They propose a multi-step prompting framework that verbalizes a social situation from multiple perspectives before forming a judgment.
Outcome: The proposed framework improves the alignment with human judgments by up to 11 F1 points with the GPT-3.5 model.
Reference Attack: A New Cross-Modal Jailbreaking Attack against Multimodal Large Language Models (2026.acl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have raised significant safety concerns about generated content, drawing attention from both academia and industry.
Approach: They propose a reference-guided cross-modal jailbreak method that enhances existing prompt-to-image injection attacks by exploiting MLLMs’ semantic reconstruction capabilities.
Outcome: The proposed method achieves an attack success rate of over 93% on leading MLLMs including ChatGPT, Gemini, Claude, and the widely used open-source LLaMA model.
“I See What You Did There”: Can Large Vision-Language Models Understand Multimodal Puns? (2026.acl-long)

Copied to clipboard

Challenge: Puns are a common form of rhetorical wordplay that exploits polysemy and phonetic similarity to create humor.
Approach: They propose a multimodal pun generation pipeline and a model to evaluate their understanding of puns.
Outcome: The proposed benchmark improves the understanding of multimodal puns by 16.5% in the F1 test.
P4: Plug-and-Play Discrete Prompting for Large Language Models Personalization (2024.findings-acl)

Copied to clipboard

Challenge: Large Language Models (LLMs) exhibit impressive capabilities in following instructions, but manually prompting them to exhibit certain personalities may result in sub-optimal performance.
Approach: They propose a plug-and-play prompting method to manipulate Large Language Models with distinct human-like personality traits by appending discrete personalized suffixes to query or dialog histories and focusing exclusively on influential tokens.
Outcome: The proposed method outperforms other prompting methods and model editing methods on four models ranging from 1.1B to 13B and achieves 79.9% accuracy in customizing LLMs’ personalities.
Just Ask One More Time! Self-Agreement Improves Reasoning of Language Models in (Almost) All Scenarios (2024.findings-acl)

Copied to clipboard

Challenge: chain-of-thought (CoT) prompting has been shown to be effective on complex reasoning tasks, but the naive greedy decoding used in CoT prompting causes the repetitiveness and local optimality.
Approach: They propose a generalizable ensemble-optimization method that uses a set of reasoning paths to prompt a language model one more time to determine the optimal answer.
Outcome: The proposed method can be generalized to almost all scenarios where the type of input questions and answer format of reasoning paths may be unknown.
Unlocking Exploration in RLVR: Uncertainty-aware Advantage Shaping for Deeper Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Reinforcement Learning with Verifiable Rewards (RLVR) has shown significant promise for enhancing the reasoning capabilities of large language models (LLMs).
Approach: They propose a model-free method that refines credit assignment by leveraging the model's internal uncertainty signals.
Outcome: Extensive experiments on five mathematical reasoning benchmarks show that the proposed method outperforms strong RLVR baselines on multiple model scales, including 1.5B and 7B.
GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick (2024.acl-long)

Copied to clipboard

Challenge: Large language models generate human-like content, but they also pose a problem with generation diversity, negatively impacting generation diversity and user experience.
Approach: They propose a Logits-Addition watermark and three variants that aim to enhance diversity to overcome generation diversity challenges.
Outcome: The Logits-Addition watermark outperforms the Logits+Trick-based watermark in diversity tests and outperformed other decoding-based methods by 0.1 to 0.3.

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