Miao Xiong

Researcher@Amazon·美国·OpenReview

研究方向

LLM Pretraining · Multimodal LLMs · Vision-Language Reasoning · Agentic AI · LLM Factuality and Interpretability · Uncertainty estimation · Calibration · Trustworthy AI

7.3

Bring Reason to Vision: Understanding Perception and Reasoning through Model Merging

ICML 2025Poster

6.0

Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs

ICLR 2024Poster

一作

5.5

$\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition

合作者 (20)

Miao Xiong

ConfTuner: Training Large Language Models to Express Their Confidence Verbally

FlipAttack: Jailbreak LLMs via Flipping

Do LLMs estimate uncertainty well in instruction-following?

FlipAttack: Jailbreak LLMs via Flipping

Bring Reason to Vision: Understanding Perception and Reasoning through Model Merging

Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs

$\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition