Jiahui Gao

PhD student@The University of Hong Kong·中国香港·OpenReview

研究方向

Multi-modal large language model · Data Distillation · Data generation · Pretrained/large language models · Zero-shot/few-shot prompting · Zero-shot/few-shot learning · Prompt learning · Few-Shot Learning · Natural Language Understanding

4.5

DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-size Canvas

ICLR 2026Poster

4.0

Look Less, Reason More: Rollout-Guided Adaptive Pixel-Space Reasoning

ICLR 2026Withdrawn

三作

7.8

TreeSynth: Synthesizing Diverse Data from Scratch via Tree-Guided Subspace Partitioning

NeurIPS 2025Spotlight

6.3

Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning

ICLR 2025Poster

二作

6.0

Forewarned is Forearmed: Harnessing LLMs for Data Synthesis via Failure-induced Exploration

ICLR 2025Poster

二作

6.0

Mixture of insighTful Experts (MoTE): The Synergy of Thought Chains and Expert Mixtures in Safety Self-Alignment

合作者 (20)

Jiahui Gao

DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-size Canvas

Look Less, Reason More: Rollout-Guided Adaptive Pixel-Space Reasoning

TreeSynth: Synthesizing Diverse Data from Scratch via Tree-Guided Subspace Partitioning

Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning

Forewarned is Forearmed: Harnessing LLMs for Data Synthesis via Failure-induced Exploration

ProReason: Multi-Modal Proactive Reasoning with Decoupled Eyesight and Wisdom

Jailbreaking as a Reward Misspecification Problem

G-LLaVA: Solving Geometric Problem with Multi-Modal Large Language Model

Implicit Search via Discrete Diffusion: A Study on Chess

Mixture of insighTful Experts (MoTE): The Synergy of Thought Chains and Expert Mixtures in Safety Self-Alignment