Tenny Yin
@tennyyin
Robotics PhD @Princeton, prev BS @Cornell
🔎Can robots search for objects like humans? Humans explore unseen environments intelligently—using prior knowledge to actively seek information and guide search. But can robots do the same? 👀 🚀Introducing WoMAP (World Models for Active Perception): a novel framework for…
🦆🚀QuACK🦆🚀: new SOL mem-bound kernel library without a single line of CUDA C++ all straight in Python thanks to CuTe-DSL. On H100 with 3TB/s, it performs 33%-50% faster than highly optimized libraries like PyTorch's torch.compile and Liger. 🤯 With @tedzadouri and @tri_dao
Reasoning makes LLMs more accurate—but also less trustworthy. We show that deeper reasoning often amplifies overconfidence, revealing a fundamental tension between accuracy and uncertainty estimation in LLMs. Check out our paper below! 👇
🤖💭Reasoning models are remarkably good at solving complex tasks through deep thinking. But do they know when they are wrong? 🧐 We show that reasoning models: ⚠️ Are typically overconfident, 🤔 Grow more overconfident with more reasoning, ⚙️ Become more trustworthy with…
Introducing WoMAP—a new active perception framework merging world models with VLMs for open-vocabulary object localization in unknown environments. Grateful to have contributed a small part.
🔎Can robots search for objects like humans? Humans explore unseen environments intelligently—using prior knowledge to actively seek information and guide search. But can robots do the same? 👀 🚀Introducing WoMAP (World Models for Active Perception): a novel framework for…
Robots struggle to find objects like humans. Why? Understanding "behind the box" (semantics) isn't enough – they need to plan precise, efficient actions to get there. Key Insight: VLMs propose where to look ("Maybe behind the box?"). World models evaluate VLM proposals and…
🔎Can robots search for objects like humans? Humans explore unseen environments intelligently—using prior knowledge to actively seek information and guide search. But can robots do the same? 👀 🚀Introducing WoMAP (World Models for Active Perception): a novel framework for…
Can we train robots to 𝒑𝒍𝒂𝒏 𝒘𝒉𝒆𝒓𝒆 𝒕𝒐 𝒍𝒐𝒐𝒌 to find an object described with language? Check out our new paper that combines prior knowledge from a VLM with a world modeling approach for planning: 🌐 WoMAP: World Models For Embodied Open-Vocabulary Object…
🔎Can robots search for objects like humans? Humans explore unseen environments intelligently—using prior knowledge to actively seek information and guide search. But can robots do the same? 👀 🚀Introducing WoMAP (World Models for Active Perception): a novel framework for…
Say ahoy to 𝚂𝙰𝙸𝙻𝙾𝚁⛵: a new paradigm of *learning to search* from demonstrations, enabling test-time reasoning about how to recover from mistakes w/o any additional human feedback! 𝚂𝙰𝙸𝙻𝙾𝚁 ⛵ out-performs Diffusion Policies trained via behavioral cloning on 5-10x data!
Want your imitation learning policy to generalize better, but how to collect data to achieve this? 🤖🤔 Enter Factored Scaling Curves (FSC): a tool that quantifies how policy success scales with demos for each environmental factor, enabling principled data collection 📈 . 🌐…