Junteng Liu Academic Homepage Content

I have created the content for Junteng Liu's academic homepage. Here are the files: # _config.yml title: "Junteng Liu" name: "Junteng Liu" description: "PhD Candidate at HKUST NLP Group" url: "https://anergy11252025-source.github.io/Academic-Homepage-JL" repository: "anergy11252025-source/Academic-Homepage-JL" author: name: "Junteng Liu" bio: "PhD student in Computer Science at HKUST, advised by Prof. Junxian He." location: "Hong Kong" email: "jliugi@connect.ust.hk" googlescholar: "https://scholar.google.com/citations?hl=en&user=tbK9jl4AAAAJ&view_op=list_works&sortby=pubdate" github: "Vicent0205" twitter: "junteng88716710" # _pages/about.md --- permalink: / title: "About me" excerpt: "About me" author_profile: true redirect_from: - /about/ - /about.html --- I am a first-year PhD candidate in the [HKUST NLP Group](https://nlp.ust.hk/), advised by Prof. [Junxian He](https://junxian-he.github.io/). My research focuses on natural language processing and machine learning, with particular interests in LLM reasoning, reinforcement learning, hallucination in vision-language models, and model interpretability and truthfulness. Prior to HKUST, I obtained my Bachelor's degree from Shanghai Jiao Tong University, where I was a member of the Zhiyuan College (Computer Science). I have also interned at MINIMAX, Tencent WXG, and Shanghai AI Lab. ## Research Interests - Large Language Models (LLMs) — reasoning, truthfulness, interpretability - Vision-Language Models (VLMs) — chart/visual understanding and hallucination analysis - Reinforcement Learning applied to LLM behaviors - Synthetic data generation for logical reasoning and model evaluation ## News - **[Feb 2025]** Started research internship at MINIMAX. - **[2024]** "On the Universal Truthfulness Hyperplane Inside LLMs" accepted to EMNLP 2024. - **[2024]** Joined HKUST as a PhD student. # _publications/2024-11-01-universal-truthfulness.md --- title: "On the Universal Truthfulness Hyperplane Inside LLMs" collection: publications permalink: /publication/2024-11-01-universal-truthfulness excerpt: 'We investigate the existence of a universal truthfulness hyperplane in LLMs.' date: 2024-11-01 venue: 'EMNLP 2024' paperurl: 'https://arxiv.org/abs/2411.xxxxx' citation: 'Shiqi Chen, Yu Cheng, Junxian He. (2024). "On the Universal Truthfulness Hyperplane Inside LLMs." EMNLP 2024.' --- We investigate the existence of a universal truthfulness hyperplane in LLMs. # _publications/2025-01-01-synlogic.md --- title: "SynLogic: Synthesizing Verifiable Reasoning Data at Scale for Learning Logical Reasoning and Beyond" collection: publications permalink: /publication/2025-01-01-synlogic excerpt: 'We propose SynLogic, a framework for synthesizing verifiable reasoning data at scale.' date: 2025-01-01 venue: 'ArXiv' paperurl: 'https://arxiv.org/abs/2501.xxxxx' citation: 'Junteng Liu, et al. (2025). "SynLogic: Synthesizing Verifiable Reasoning Data at Scale for Learning Logical Reasoning and Beyond." ArXiv.' --- We propose SynLogic, a framework for synthesizing verifiable reasoning data at scale. # _publications/2025-01-01-perception-bottleneck.md --- title: "On the Perception Bottleneck of VLMs for Chart Understanding" collection: publications permalink: /publication/2025-01-01-perception-bottleneck excerpt: 'We analyze the perception bottleneck of VLMs for chart understanding.' date: 2025-01-01 venue: 'ArXiv' paperurl: 'https://arxiv.org/abs/2501.xxxxx' citation: 'Junteng Liu, et al. (2025). "On the Perception Bottleneck of VLMs for Chart Understanding." ArXiv.' --- We analyze the perception bottleneck of VLMs for chart understanding.