NTU
academicNanyang Technological University (est. 1991) is the academic engine behind much of the open large-multimodal-model ecosystem, driven almost entirely by Ziwei Liu (MMLab@NTU, CCDS) and the LMMs-Lab community he anchors with Bo Li. An organizational note matters: LMMs-Lab (EvolvingLMMs-Lab on GitHub, lmms-lab on HuggingFace) is formally an independent open research community, not an NTU unit — NTU researchers anchor it, so attribution here is per-item; the MMLab-NTU GitHub org itself is empty.
Flagship lines: LMMs-Eval, the de-facto standard multimodal evaluation harness (4.3K★); the LLaVA-OneVision family co-developed with ByteDance (0.5B–72B, 220K+ monthly downloads on the 7B), evolving into the fully open LLaVA-OneVision-1.5 ($16K training budget) and LLaVA-OneVision-2; the EgoLife/EgoGPT egocentric assistant project (CVPR 2025); the Video-MMMU benchmark; and the legacy Otter that started the lineage in 2023 (slug otter-lmm — the bare slug belongs to Cambridge's Otter weather model).
Chen Change Loy's S-Lab runs restoration and generation (SeedVR/SeedVR2 with ByteDance Seed; Light-X 4D video rendering at ICLR 2026), and Bo An ships LLM-agent infrastructure (AgentOrchestra, with Skywork). Attribution flags: the Vchitect video-generation stack sits under Shanghai AI Lab, and ByteDance-led joint work (MMSearch-R1, SeedVR) files with ByteDance Seed first.
Your notes
People
- Ziwei Liu Google Scholar — Professor (Provost's Chair in AI), CCDS; leads MMLab@NTU and anchors LMMs-Lab
- Chen Change Loy Google Scholar — President's Chair Professor; leads S-Lab (SeedVR line); CVPR 2026 Program Co-Chair
- Bo An Google Scholar — President's Chair Professor, CCDS; multi-agent RL and LLM-agent infrastructure (AgentOrchestra)
- Shijian Lu Google Scholar — Associate Professor; vision-language and 3D scene understanding
- Erik Cambria Google Scholar — Professor; SenticNet and affective computing