Publications

My research centers on building trustworthy knowledge-intensive AI systems, organized around a unified question-answering pipeline: given a user query, (1) understanding what the query is truly asking, (2) assessing whether the model knows the answer, (3) evaluating whether external evidence is reliable, and (4) effectively leveraging external knowledge when needed. Each stage addresses a critical challenge in ensuring that AI systems produce accurate, honest, and well-grounded responses.

Research overviewKnowledge Boundaries & Honest AI Work in ProgressA connected story of our work on self-assessment, calibration, and honesty alignment.

What is the Query asking for?

Clarifying Question Generation / Facet Generation

SIGIR-AP'23

A Comparative Study of Training Objectives for Clarification Facet Generation

Shiyu Ni, Keping Bi, Jiafeng Guo, and Xueqi Cheng

Does the Model Know the Answer?

LLM Honesty Alignment / Self-assessment

Preprint

Shared Experience, Separate Learning: Companion Confidence Calibration for LLMs

Shiyu Ni, Keping Bi, Jiafeng Guo, Yilong Xu, Jingtong Wu, Zengxin Han, and Xueqi Cheng

EMNLP'26 Oral

Popular but Wrong: Understanding and Mitigating LLM Overconfidence through Knowledge Popularity

Shiyu Ni, Keping Bi, Jiafeng Guo, and Xueqi Cheng

ICLR'26

Annotation-Efficient Universal Honesty Alignment

Shiyu Ni, Keping Bi, Jiafeng Guo, Minghao Tang, Jingtong Wu, Zengxin Han, and Xueqi Cheng

ACL'25

Towards Fully Exploiting LLM Internal States to Enhance Knowledge Boundary Perception

Shiyu Ni, Keping Bi, Jiafeng Guo, Lulu Yu, Baolong Bi, and Xueqi Cheng

ACL'24 Findings

When Do LLMs Need Retrieval Augmentation? Mitigating LLMs' Overconfidence Helps Retrieval Augmentation

Shiyu Ni, Keping Bi, Jiafeng Guo, and Xueqi Cheng

CCIR'24

Are Large Language Models More Honest in Their Probabilistic or Verbalized Confidence?

Shiyu Ni, Keping Bi, Lulu Yu, and Jiafeng Guo

EMNLP'25 Findings

Do LVLMs Know What They Know? A Systematic Study of Knowledge Boundary Perception in LVLMs

Zhikai Ding, Shiyu Ni, and Keping Bi

arXiv

Evaluating and Calibrating LLM Confidence on Questions with Multiple Correct Answers

Yuhan Wang^†, Shiyu Ni^†, Zhikai Ding, Zihang Zhan, Yuanzi Li, and Keping Bi

LLM-as-a-judge / External Verifier

ACL'26 · VulGen@SIGIR2026 SAC Oral Recommendation

How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality

Minzhu Tu^†, Shiyu Ni^†, and Keping Bi

Is External Evidence Good?

Query Performance Prediction

arXiv

Can LLM Rerankers Predict Their Own Ranking Performance?

Shiyu Ni, Keping Bi, Jiafeng Guo, Jingtong Wu, Zengxin Han, and Xueqi Cheng

How to Leverage External Knowledge?

Interact Using Tokens

SIGIR-AP'25

Injecting External Knowledge into the Reasoning Process Enhances Retrieval-Augmented Generation

Minghao Tang, Shiyu Ni, Jiafeng Guo, and Keping Bi

Parametric Injection

arXiv

The Role of Parametric Injection-A Systematic Study of Parametric Retrieval-Augmented Generation

Minghao Tang, Shiyu Ni, Jingtong Wu, Zengxin Han, and Keping Bi

Collaborations

ICLR'25

Is Factuality Enhancement a Free Lunch For LLMs? Better Factuality Can Lead to Worse Context-Faithfulness

Baolong Bi, Shenghua Liu, Yiwei Wang, Lingrui Mei, Junfeng Fang, Hongcheng Gao, Shiyu Ni, and Xueqi Cheng

arXiv

Deep Research: A Systematic Survey

Zhengliang Shi, Yiqun Chen, Haitao Li, Weiwei Sun, Shiyu Ni, Yougang Lyu, Runze Fan, et al.

^† Equal contribution.