LMMs-Lab
Feeling and building multimodal intelligence.
Explained projects
Yet to be explained
OneVision-Encoder
Explain
Codec-Aligned Sparsity as a Foundational Principle for Multimodal Intelligence
lmms-lab-writer
Explain
Agentic LaTeX Writer - Local-first editor for AI-assisted academic writing
LongVT
Explain
[CVPR 2026] LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling
SkillOpt-Lite
Explain
SkillOpt-Lite and HarnessOpt: Optimize your skill or harness with one line of vibe
OpenMMReasoner
Explain
[CVPR 2026] OpenMMReasoner: Pushing the Frontiers for Multimodal Reasoning with an Open and General Recipe
Evolving-Visual-Generation
Explain
[Roadmap] Visual Generation in the New Era: An Evolution from Atomic Mapping to Agentic World Modeling
ParaVT
Explain
[NeurIPS 2026] ParaVT: Taming the Tool Prior Paradox for Parallel Tool Use in Agentic Video Reinforcement Learning
LLaVA-OneVision-1.5-RL
Explain
Fully Open Framework for Democratized Multimodal Reinforcement Learning.
GenAI-Caption-Pipeline
Explain
Reconstructing Big Tech T2I/T2V captioning pipelines for production and research
MacOS-H3-Speedrun
Explain
Full MiniMax H3 speedrun workflows for Apple Silicon with strict staging, Turbo, FBC, Spectrum, and SolAttn
engram
Explain
Privacy-first AI memory layer - Signal for AI Memory. E2EE, local-first, works with Claude, Cursor, and any MCP-compatible AI.
VLMEvalKit
Explain
An open-source evaluation toolkit to evaluate MLLMs on Spatial Intelligence using the EASI protocol
co-scientist
Explain
For AI Agents to post ideas on their owns.
symbols-or-pixels
Explain
Code and per-sample results for the LMMs-Lab note "Symbols or Pixels?": zero-shot language models versus VBVR-Pro video models on VBVR-Pro-Bench.
homebrew-tap
Explain
Homebrew tap for LMMs-Lab applications