DPhil Student · University of Oxford

Runjia Li

I work on consistent, physically-plausible, and interactive video generation.

I am a DPhil student at the University of Oxford, supervised by Andrea Vedaldi and Philip Torr, and generously funded by the Clarendon Scholarship. Before that I completed my undergraduate studies in Engineering Science, also at Oxford. I am fortunate to collaborate with Tomas Jakab.

Experience

Meta2026

Research Scientist Intern

Snap Inc.2025

Research Scientist Intern

Publications

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

A dataset, a real-time streaming model, and a benchmark for instruction-based editing of egocentric video.

Dataset

VMem: Consistent Interactive Video Scene Generation with Surfel-Indexed View Memory

Runjia Li, Philip Torr, Andrea Vedaldi, Tomas Jakab

ICCV 2025 Highlight

Interactive scene generation that stays consistent as the camera revisits places, by indexing past views with surfels and recalling only the ones a new view actually needs.

Project Page arXiv GitHub

Demo

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

Generating 3D fantastical animals with distinct, correctly placed parts, by transferring part-aware knowledge from a 2D diffusion model into the 3D distillation process.

Project Page arXiv GitHub

CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor

Runjia Li*, Shuyang Sun*, Philip Torr, Xiuye Gu, Siyang Li

CVPR 2024

Training-free open-vocabulary segmentation: querying CLIP recurrently prunes away text prompts that are absent from the image and sharpens the masks for the ones that remain.

Project Page arXiv GitHub

OxfordTVG-HIC: Can Machine Make Humorous Captions from Images?

Runjia Li*, Shuyang Sun*, Mohamed Elhoseiny, Philip Torr

ICCV 2023

A large-scale dataset of humorous image captions, together with an evaluation of how well models can generate and judge humour.

Project Page arXiv GitHub