🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Learn more about the state-of-the-art in 3D Gaussian Splatting (3DGS) compression and how compression differs from compaction.

Exciting Opportunity! Wuyang Chen , Assistant Professor at SFU, is recruiting PhD students for Fall 2025 in AI/ML! 🚀 🌐 Explore thier research: Delta Lab (delta-lab-ai.github.io/index.html) 📬 Interested? Reach out to Wuyang & apply! #AI #MachineLearning #PhDOpportunity #SFU

An underrated but occasionally make-or-break skill in AI research (that didn’t really exist ten years ago) is the ability to find a dataset that actually exercises a new method you are working on. Back in the day when the bottleneck in AI was learning, many methods were

🚀 Our latest work, VideoLifter, is here! VideoLifter converts long video sequences into 3D representations with unprecedented speed, achieving 82% faster training while maintaining state-of-the-art accuracy. 🔗 Explore the details and visuals on our project page:

Picture a ball flying toward you: you don’t replay every frame to know it’ll hit. Today’s world models predict visually, understand via captions, and logic through text. The next leap? World Models that fuse neural states with symbolic logic for human-like reasoning.

🎶 Excited to introduce SPIN-Bench! 🎉 TL;DR: Benchmarking LLM capabilities across various strategic planning game environments. 🚀 🌐 Project Page: spinbench.github.io 📄 arXiv: arxiv.org/abs/2503.12349 🎮 Interact with PDDL domains: spinbench.github.io/tools/pddl/tra… 📉 LLM

Claude-sonnet (White) vs. o1-mini (Black): Given an absolutely winning poisoned position, o1-mini boldly plays... e5h5! 🤔 Small legal action space—but vast stage action space—places heavy cognitive burden on LLMs. Explore more chess examples & insights in SPIN-bench here:

How well Do LLMs Plan Strategically? Can they beat Humans in board games? SPIN or Strategic Planning, Interaction, and Negotiation is a new multi-domain evaluation showing LLMs can do basic planning, but fail in complex strategic and social reasoning tasks compared to humans.

ReLU, but for geopolitics

One of the best 3D vision talks I've seen recently. Clear and insightful.

🚀 Proud to share our new arXiv preprint: VLM-3R learns 3D spatial-temporal reasoning straight from monocular video—no depth sensors or prebuilt maps! Trained on 200K+ 3D QA instructs 👉 vlm-3r.github.io