RAVEN is open-source: code, paper, and RAVEN-QA benchmark below.
We'll be at the SISREN Workshop (#CoRL2026) in Austin on Nov 12. Working on robot memory or embodied navigation? Come say hi! 🤖
🌐 https://t.co/heTy8pWEyI
💻 https://t.co/st9Mb7XScK
📄 https://t.co/AnGE1UPdAW
How should a robot remember what it saw weeks ago: a blue hat with a pink bow, a fire extinguisher, where you left your keys?
Today's robot memories caption everything into text first, losing the visual details that make retrieval work.
Please meet RAVEN: a long-horizon memory that stores visual embeddings directly — no captioning bottleneck. It answers questions and navigates to goals across days of experience.
🎉 Accepted to the SISREN Workshop at #CoRL2026! 🧵👇
https://t.co/gq8810eNI2
RAVEN scales to long horizons.
On FindingDory Long-Horizon episodes averaging ~3,000 frames, RAVEN surpasses querying a frontier VLM directly, while retrieving only ~7.4% of memory per query, making it 13.5× more memory-efficient.
Meet KinDER — a stress test for robot physical reasoning. All 13 methods failed 😈
🌎 25 environments
♾️ Infinite tasks
🏋️ Gymnasium API
⚒️ Over 20 parameterized skills
🪧 Human demonstrations
📊 13 baselines (planning and learning)
From @Princeton@CMU_Robotics@ICatGT@CambridgeMLG@nvidia@MIT_CSAIL
🧵 1/n
New Meta Research 🚀
AI agents are powerful, but don’t stay aligned with you over time.
When preferences shift, they don’t adapt. You correct them once…they repeat the mistake. 🤦
Introducing PAHF: continual personalization where agents learn from feedback to stay in sync.