All News

LEGO-Puzzles is published at ECCV 2026.

Two papers are accepted by EMNLP 2026, studying multimodal retrieval heads and controllable non-Markov games.

Seed2.1 is released, advancing multimodal reasoning and agentic productivity across the Seed model family.

Five papers are published at CVPR 2026, covering GUI-agent evaluation, spatial self-supervised RL, agentic reward modeling, visual reasoning, and poster intelligence.

Two papers are published at ACL 2026: OPT-BENCH and NP-Engine.

Seed2.0 is officially launched, strengthening multimodal understanding, complex instruction execution, and real-world agent capabilities.

Three papers are published at ICLR 2026: MMSI-Bench, VisualPRM, and MM-HELIX.

Seed1.8 is officially released as a generalized agentic model with multimodal, search, coding, and GUI capabilities.

I joined ByteDance Seed in Singapore, after two years at Shanghai AI Laboratory working on OpenCompass and large-model evaluation.

RISEBench is accepted by NeurIPS 2025 Datasets & Benchmarks as an Oral presentation.

InternVL3.5 is released with stronger reasoning, efficiency, and deployment support.

Four papers are published at ICCV 2025, spanning visual reinforcement fine-tuning, instruction following, creative intelligence, and benchmark information density.

Visual-RFT is accepted by ICCV 2025.

Five papers are published at ACL 2025, including OmniAlign-V, Redundancy Principles, and Condor.

We release MMSI-Bench, a benchmark for multi-image spatial intelligence.

We release RISEBench for reasoning-informed visual editing evaluation.

We release LEGO-Puzzles, a benchmark for multi-step spatial reasoning in MLLMs.

Six papers are accepted by NeurIPS 2024: three main-conference papers (InternLM-XComposer2-4KHD, MMStar, and Prism) and three Datasets & Benchmarks papers (ShareGPT4Video, GMAI-MMBench, and MMBench-Video).

MMBench is accepted by ECCV 2024 as an Oral presentation.

VLMEvalKit is accepted by ACM MM 2024.

MathBench is accepted by ACL 2024 Findings.

Two papers are accepted by NAACL 2024: BotChat and Ada-LEval.

We release VLMEvalKit, an all-in-one toolkit for evaluating large multimodal models.

SkeleTR is published at ICCV 2023, extending skeleton-based action recognition to in-the-wild settings.

I received my Ph.D. from MMLab @ CUHK and joined Shanghai AI Laboratory as a postdoctoral researcher.

We release MMBench, a systematic benchmark for measuring the all-around capabilities of multimodal models.

JourneyDB is published at the NeurIPS 2023 Datasets & Benchmarks Track.

We release PYSKL, a codebase for skeleton action recognition; the system paper is later accepted by ACM MM 2022.

Three papers are accepted by CVPR 2022: PoseC3D (Oral), TransRank (Oral), and OCSampler.

OmniSource is published at ECCV 2020, studying omni-sourced webly supervised learning for video recognition.

MMAction2 becomes OpenMMLab's modular toolbox and benchmark for video understanding.

TRB, a triplet representation for understanding 2D human bodies, is published at ICCV 2019.

I received my B.S. in Data Science from Peking University and began my Ph.D. at The Chinese University of Hong Kong.