Publications

AudioTrust: Benchmarking the Multifaceted Trustworthiness of Audio Large Language Models

ICLR 2026

Authors: Kai Li, Can Shen, Yile Liu, et al., Hanjun Luo, others, Xinfeng Li.

Insight: We built a multifaceted benchmark and evaluation pipeline to measure reliability, safety, and robustness consistently for audio LLMs. Trustworthy audio LLM evaluation should jointly cover these dimensions instead of testing only one axis.

PrefIx: Understand and Adapt to User Preference in Human-Agent Interaction

arXiv 2026

Authors: Jialin Li, Zhenhao Chen, Hanjun Luo, Hanan Salam.

Insight: We designed a framework that infers and updates user preference signals during interaction. Preference adaptation should be modeled continuously during multi-turn human-agent collaboration.

HAI-Eval: Measuring Human-AI Synergy in Collaborative Coding

arXiv 2025

Authors: Hanjun Luo, Chiming Ni, Jiaheng Wen, et al., Hanan Salam.

Insight: We introduced a collaborative coding evaluation setup to quantify human-AI synergy end to end. Human-AI coding quality should be assessed through joint performance rather than isolated model generations.

Agentauditor: Human-level Safety and Security Evaluation for LLM Agents

NeurIPS 2025

Authors: Hanjun Luo, Shenyu Dai, Chiming Ni, et al., Hanan Salam.

Insight: We proposed AgentAuditor to test realistic agent behaviors across safety and security risk scenarios. Agent safety must be evaluated as an end-to-end system, not only by isolated model outputs.

DynamicNER: A Dynamic, Multilingual, and Fine-Grained Dataset for LLM-based Named Entity Recognition

EMNLP 2025 Main

Authors: Hanjun Luo, Yingbin Jin, Yiran Wang, et al., Zuozhu Liu.

Insight: We built DynamicNER with new data construction and evaluation protocols tailored to LLM-based NER. NER evaluation for LLMs benefits from dynamic, multilingual, and fine-grained settings beyond static benchmarks.

A Comprehensive Survey in LLM(-agent) Full Stack Safety: Data, Training and Deployment

arXiv 2025

Authors: Kun Wang, Guibin Zhang, Zhenhong Zhou, et al., Hanjun Luo, others, Yang Liu.

Insight: We organized a full-stack taxonomy and synthesized mitigation strategies across the LLM/agent lifecycle. Safety risks propagate across the whole stack, so mitigation must align data, training, and deployment stages.

AutoDebias: Automated Framework for Debiasing Text-to-Image Models

arXiv 2025

Authors: Hongyi Cai, Mohammad Mahdinur Rahman, Mingkang Dong, et al., Hanjun Luo, Yang Liu.

Insight: We implemented an automated framework that generates and applies debiasing interventions for text-to-image models. Automated debiasing pipelines can reduce social bias without full model retraining.

BIGbench: A Unified Benchmark for Evaluating Multi-dimensional Social Biases in Text-to-Image Models

arXiv 2024

Authors: Hanjun Luo, Haoyu Huang, Ziye Deng, et al., Zuozhu Liu.

Insight: We built BIGbench to provide a unified dataset and metric suite for multi-dimensional bias evaluation. Bias in text-to-image systems is multi-dimensional and needs unified metrics instead of single-axis checks.

Faintbench: A Holistic and Precise Benchmark for Bias Evaluation in Text-to-Image Models

arXiv 2024

Authors: Hanjun Luo, Ziye Deng, Ruizhe Chen, Zuozhu Liu.

Insight: We developed FAIntbench with a holistic task design and more precise bias assessment criteria. Fine-grained bias auditing needs both semantic coverage and precise scoring protocols to avoid misleading conclusions.

VersusDebias: Universal Zero-shot Debiasing for Text-to-Image Models via SLM-based Prompt Engineering and Generative Adversary

arXiv 2024

Authors: Hanjun Luo, Ziye Deng, Haoyu Huang, Xuecheng Liu, Ruizhe Chen, Zuozhu Liu.

Insight: We proposed VersusDebias, an SLM-guided prompting and adversarial strategy for practical debiasing. Zero-shot debiasing can be generalized with adaptive prompting without retraining diffusion backbones.

Uniap: Towards Universal Animal Perception in Vision via Few-shot Learning

AAAI 2024

Authors: Meiqi Sun, Zhonghan Zhao, Wenhao Chai, Hanjun Luo, et al., Gaoang Wang.

Insight: We implemented and validated a universal animal perception framework across diverse vision benchmarks. Few-shot transfer can unify multiple animal perception tasks under one scalable vision framework.