Junlin Han

Member of Technical Staff, AMI Labs

Email: junlinhcv@gmail.com · junlin@amilabs.xyz · junlin.han@eng.ox.ac.uk

GitHub CV Google Scholar Twitter/X

Profile

Welcome to my website.

I am a Member of Technical Staff at AMI Labs, where I work on pre-training and the capabilities of world models.
I did my PhD in the Torr Vision Group at the University of Oxford, advised by Prof. Philip Torr. During my PhD, I spent three years at Meta (FAIR and GenAI) as a joint PhD student working on multimodal pre-training, MLLMs, and 3D/video generation with Dr. Mike Lewis and Dr. Filippos Kokkinos.
I received my BSc from the Australian National University, where I was fortunate to be supervised by Dr. Lars Petersson, Prof. Hongdong Li, and Prof. Ian Reid.

News

[10.2026] Joined AMI Labs as a Member of Technical Staff.
[06.2026] Two papers were accepted to ECCV 2026: Tempo (long video understanding) and automatic red teaming of text-to-image models.
[06.2026] Blog post: 3D is not intelligence. If we aim for intelligence, 3D research should move beyond single modality and embrace unified models.
[03.2026] Released a large-scale exploration of unified multimodal pre-training, with practical insights into the next stage of pre-training.
[02.2026] Two papers were accepted to CVPR 2026: EgoEdit and AuralSAM2.
[01.2026] Two papers were accepted to ICLR 2026: Learning to See Before Seeing (oral) and Kaleido.
[11.2025] Joined FAIR to work with Mike Lewis on unified MLLM pre-training.
[09.2025] Hallucination at a Glance was accepted to NeurIPS 2025.
[07.2025] 3D2EP was accepted to SIGGRAPH 2025.
[05.2025] Flex3D was accepted to ICML 2025.
[03.2025] A paper on point-cloud segmentation was accepted to CVPR 2025.
[12.2024] Our previous work on low-level vision received the Best Paper Award in Remote Sensing.
[12.2024] We organized the 2nd Workshop on Foundation Models in the Wild at ICLR 2025.
[11.2024] Two papers, 3D-GPT and DreamBeast, were accepted to 3DV 2025.
[07.2024] VFusion3D and Unicorns were accepted to ECCV 2024.

Show more ▼

Research Interests

My research focuses on multimodal foundation models, including multimodal large language models and generative models, as well as their interactions.

I am particularly interested in multimodal approaches to advanced intelligence, especially how vision can contribute to its development through large-scale pre-training.

Multimodal pre-training

Selected projects: Physics of multimodal pretraining, Visual priors in LLM pre-training, Beyond language modeling

Multimodal post-training

Selected projects: Llama 4, Long video understanding

Generative models, especially 3D and video

Selected projects: Scaling up 3D generation, High-quality 3D asset generation, Scaling up neural rendering

Education

DPhil in Computer Vision, University of Oxford

January 2024 – Present

Supervisor: Prof. Philip Torr

BSc in Information Technology (Honours), Australian National University

February 2019 – July 2023

Supervisors: Dr. Lars Petersson, Prof. Hongdong Li, and Prof. Ian Reid

Experience

Member of Technical Staff, AMI Labs

October 2026 – Present

Pre-training and capabilities of world models

PhD Student Researcher, Meta

October 2023 – September 2026

FAIR (December 2025 – September 2026), hosted by Dr. Mike Lewis — multimodal pre-training
GenAI Llama Vision (August 2024 – December 2025), hosted by Dr. Filippos Kokkinos — MLLM post-training and pre-training
GenAI RepliGen (October 2023 – July 2024), hosted by Dr. Filippos Kokkinos — 3D and video generation

Publications

Recent representative papers are highlighted.

	Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes Junlin Han, Shengbang Tong, David Fan, Minghao Chen, Philip Torr, Filippos Kokkinos, Mike Lewis Preprint, 2026 [Project page] [arXiv] [PDF]
	Beyond Language Modeling: An Exploration of Multimodal Pretraining Shengbang Tong, David Fan, John Nguyen, Ellis Brown, Gaoyue Zhou, Shengyi Qian, Boyang Zheng, Théophane Vallaeys, Junlin Han, Rob Fergus, Naila Murray, Marjan Ghazvininejad, Mike Lewis, Nicolas Ballas, Amir Bar, Michael Rabbat, Jakob Verbeek, Luke Zettlemoyer, Koustuv Sinha, Yann LeCun, Saining Xie ICML (spotlight, top 2.2%), 2026 [Project page] [arXiv] [PDF]
	Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos ICLR (oral, top 1.18%), 2026 [Project page] [arXiv] [PDF]
	VFusion3D: Learning Scalable 3D Generative Models from Video Diffusion Models Junlin Han, Filippos Kokkinos, Philip Torr ECCV, 2024 [Project page] [arXiv] [PDF] [Code]

	Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes Junlin Han, Shengbang Tong, David Fan, Minghao Chen, Philip Torr, Filippos Kokkinos, Mike Lewis Preprint, 2026 [Project page] [arXiv] [PDF]
	PartiCam: Camera Controlled Video Generation with Reward Guidance Amine Ouasfi, Runjia Li, Junlin Han, Eric Marchand, Philip Torr, Adnane Boukhayma Preprint, 2026 [arXiv] [PDF]
	AnimalLift: Reconstructing Animatable 3D Animals from a Single Image by Learning Canonical Shape, Texture, and Fur Maps Chunyi Sun, Ruyi Zha, Weijian Deng, Junlin Han, Dylan Campbell, Stephen Gould SIGGRAPH Asia, 2026 [arXiv] [PDF]
	Automatic Red Teaming for Implicit Vulnerabilities of Text-to-Image Models Chang Ma, Junlin Han, Shuo Chen, Runjia Li, Philip Torr, Jindong Gu ECCV, 2026 [arXiv] [PDF] [Springer]
	Small Vision-Language Models are Smart Compressors for Long Video Understanding Junjie Fei, Jun Chen, Zechun Liu, Yunyang Xiong, Chong Zhou, Wei Wen, Junlin Han, Mingchen Zhu, Saksham Suri, Qi Qian, Shuming Liu, Lemeng Wu, Raghuraman Krishnamoorthi, Vikas Chandra, Mohamed Elhoseiny, Chenchen Zhu ECCV, 2026 [Project page] [Code] [arXiv] [PDF] [Demo]
	Beyond Language Modeling: An Exploration of Multimodal Pretraining Shengbang Tong, David Fan, John Nguyen, Ellis Brown, Gaoyue Zhou, Shengyi Qian, Boyang Zheng, Théophane Vallaeys, Junlin Han, Rob Fergus, Naila Murray, Marjan Ghazvininejad, Mike Lewis, Nicolas Ballas, Amir Bar, Michael Rabbat, Jakob Verbeek, Luke Zettlemoyer, Koustuv Sinha, Yann LeCun, Saining Xie ICML (spotlight, top 2.2%), 2026 [Project page] [arXiv] [PDF]
	EgoEdit: Dataset, Real-Time Streaming Model, and Benchmark for Egocentric Video Editing Runjia Li, Moayed Haji-Ali, Ashkan Mirzaei, Chaoyang Wang, Arpit Sahni, Ivan Skorokhodov, Aliaksandr Siarohin, Tomas Jakab, Junlin Han, Sergey Tulyakov, Philip Torr, Willi Menapace CVPR, 2026 [Project page] [arXiv] [pdf]
	Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos ICLR (oral, top 1.18%), 2026 [Project page] [arXiv] [PDF]
	Scaling Sequence-to-Sequence Generative Neural Rendering Shikun Liu, Kam Woh Ng, Wonbong Jang, Jiadong Guo, Junlin Han, Haozhe Liu, Yiannis Douratsos, Juan C. Pérez, Zijian Zhou, Chi Phung, Tao Xiang, Juan-Manuel Pérez-Rúa ICLR, 2026 [Project page] [arXiv] [pdf]
	AuralSAM2: Enabling SAM2 Hear Through Pyramid Audio-Visual Feature Prompting Yuyuan Liu, Yuanhong Chen, Chong Wang, Junlin Han, Junde Wu, Can Peng, Jingkun Chen, Yu Tian, Gustavo Carneiro CVPR findings, 2026 [arXiv] [pdf]
	Learning-based Multi-View Stereo: A Survey Fangjinhua Wang, Qingtian Zhu, Di Chang, Quankai Gao, Junlin Han*, Tong Zhang, Richard Hartley, Marc Pollefeys TPAMI, 2026 [arXiv] [pdf]
	From Pixels to Feelings: Aligning MLLMs with Human Cognitive Perception of Images Yiming Chen, Junlin Han, Tianyi Bai, Shengbang Tong, Filippos Kokkinos, Philip Torr Preprint, 2025 [Project page] [arXiv] [pdf]
	Unsupervised Decomposition of 3D Shapes into Expressive and Editable Extruded Profile Primitives Chunyi Sun, Junlin Han, Runjia Li, Weijian Deng, Dylan Campbell, Stephen Gould SIGGRAPH, 2025 [paper] [pdf]
	Hallucination at a Glance: Controlled Visual Edits and Fine-Grained Multimodal Learning Tianyi Bai, Yuxuan Fan, Jiantao Qiu, Fupeng Sun, Jiayi Song, Junlin Han, Zichen Liu, Conghui He, Wentao Zhang, Binhang Yuan NeurIPS, 2025 [Dataset] [arXiv] [pdf]
	Flex3D: Feed-Forward 3D Generation with Flexible Reconstruction Model and Input View Curation Junlin Han, Jianyuan Wang, Andrea Vedaldi, Philip Torr, Filippos Kokkinos ICML, 2025 [Project page] [arXiv] [PDF]
	The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation Llama4 team Models, 2025 [project page] [Models]
	VGRP-Bench: Visual Grid Reasoning Puzzle Benchmark for Large Vision-Language Models Yufan Ren, Konstantinos Tertikas, Shalini Maiti, Junlin Han, Tong Zhang, Sabine Süsstrunk, Filippos Kokkinos Preprint, 2025 [Project page] [arXiv] [pdf] [Code] [Dataset]
	Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model Zhaochong An, Guolei Sun, Yun Liu, Runjia Li, Junlin Han, Ender Konukoglu, Serge Belongie CVPR, 2025 [arXiv] [pdf] [Code]
	3D-GPT: Procedural 3D Modeling with Large Language Models Chunyi Sun, Junlin Han, Weijian Deng, Xinlong Wang, Zishan Qin, Stephen Gould 3DV, 2025 [Project page] [arXiv] [pdf] [Code]
	DreamBeast: Distilling 3D Fantastical Animals with Part-Aware Knowledge Transfer Runjia Li, Junlin Han, Luke Melas-Kyriazi, Chunyi Sun, Zhaochong An, Zhongrui Gui, Shuyang Sun, Philip Torr, Tomas Jakab 3DV, 2025 [Project page] [arXiv] [pdf] [Code]
	VFusion3D: Learning Scalable 3D Generative Models from Video Diffusion Models Junlin Han, Filippos Kokkinos, Philip Torr ECCV, 2024 [Project page] [arXiv] [PDF] [Code]
	Semantic Score Distillation Sampling for Compositional Text-to-3D Generation Ling Yang, Zixiang Zhang, Junlin Han, Bohan Zeng, Runjia Li, Philip Torr, Wentao Zhang Preprint, 2024 [arXiv] [pdf]
	Strong and Controllable Blind Image Decomposition Zeyu Zhang, Junlin Han, Chenhui Gou*, Hongdong Li, Liang Zheng Preprint, 2024 [arXiv] [pdf]
	How Many Unicorns Are in This Image? A Safety Evaluation Benchmark for Vision LLMs Haoqin Tu, Chenhang Cui, Zijun Wang, Yiyang Zhou, Bingchen Zhao, Junlin Han*, Wangchunshu Zhou, Huaxiu Yao, Cihang Xie ECCV, 2024 [arXiv] [pdf] [Code]
	NeRFEditor: Differentiable Style Decomposition for 3D Scene Editing Chunyi Sun, Yanbin Liu, Junlin Han, Stephen Gould WACV, 2024 [Project page] [arXiv] [pdf]
	What Images are More Memorable to Machines? Junlin Han, Huangying Zhan, Jie Hong, Pengfei Fang, Hongdong Li, Lars Petersson, Ian Reid Preprint, 2023 [Project page] [arXiv] [pdf] [Code]
	Hyperbolic Audio-visual Zero-shot Learning Jie Hong, Zeeshan Hayder, Junlin Han, Pengfei Fang, Mehrtash Harandi, Lars Petersson ICCV, 2023 [arXiv] [pdf] [Code]
	Curved Geometric Networks for Visual Anomaly Recognition Jie Hong, Pengfei Fang, Weihao Li, Junlin Han, Lars Petersson, Mehrtash Harandi TNNLS, 2023 [Paper] [arXiv] [pdf] [Code]
	GOSS: Towards Generalized Open-set Semantic Segmentation Jie Hong, Weihao Li, Junlin Han, Jiyang Zheng, Pengfei Fang, Mehrtash Harandi, Lars Petersson The Visual Computer, 2023 [Paper] [arXiv] [pdf] [Code]
	You Only Cut Once: Boosting Data Augmentation with a Single Cut Junlin Han, Pengfei Fang, Weihao Li, Jie Hong, Ali Armin, Ian Reid, Lars Petersson, Hongdong Li ICML, 2022 [arXiv] [pdf] [PMLR] [Code] [Slide]
	Blind Image Decomposition Junlin Han, Weihao Li, Pengfei Fang, Chunyi Sun, Jie Hong, Ali Armin, Lars Petersson, Hongdong Li ECCV, 2022 [Project page] [arXiv] [pdf] [Code] [Dataset] [Video] [Slide] [Poster]
	CropMix: Sampling a Rich Input Distribution via Multi-Scale Cropping Junlin Han, Lars Petersson, Hongdong Li, Ian Reid Preprint, 2022 [arXiv] [pdf] [Code]
	Underwater Image Restoration via Contrastive Learning and a Real-world Dataset Junlin Han, Mehrdad Shoeiby, Tim Malthus, Elizabeth Botha, Janet Anstee, Saeed Anwar, Ran Wei, Ali Armin, Hongdong Li, Lars Petersson Remote Sensing (best paper award, top 0.1%), 2022 [Paper] [Code] [Dataset]
	Single Underwater Image Restoration by Contrastive Learning Junlin Han, Mehrdad Shoeiby, Tim Malthus, Elizabeth Botha, Janet Anstee, Saeed Anwar, Ran Wei, Lars Petersson, Ali Armin IGARSS (oral), 2021 [arXiv] [pdf] [Code] [Dataset]
	Dual Contrastive Learning for Unsupervised Image-to-Image Translation Junlin Han, Mehrdad Shoeiby, Lars Petersson, Ali Armin NTIRE, CVPRW (oral), 2021 [arXiv] [pdf] [CVF] [Code] [Video]

Blogs

From 3D Vision to Multimodal Intelligence
June 2026 — If we aim for intelligence, 3D research should move beyond single modality and embrace unified models.

Academic Services

Workshop organization: Foundation Models in the Wild at ICML 2024 and ICLR 2025
Conference reviewing: CVPR, ICCV, ECCV, ICML, NeurIPS, ICLR, COLM, SIGGRAPH Asia
Journal reviewing: TPAMI, IJCV, TIP, TGRS

Awards

Meta PhD Scholarship, 2023
Best Paper Award, 2022, Remote Sensing
Chancellor's Letter of Commendation, 2022, Australian National University
Top Reviewer, 2022, NeurIPS
Second Best Presentation Award, 2021, AIM Conference
Top-up Scholarship, 2021, Data61-CSIRO
Undergraduate Vacation Scholarship, 2020, Data61-CSIRO

Misc

I was born and raised in Chengdu, a city known for its food and pandas.
I have studied and lived in China, Singapore, Australia, and the UK.
Outside research, I enjoy nature travel, cooking, gaming (e.g., Teamfight Tactics and Hearthstone; previously ranked top 50 worldwide), and swimming.