Ask a 3D vision-language model what's near the table and in front of the curtain, and it might guess "sewing machine." The right answer is a tray rack.
CVP (UC San Diego + Lambda, WACV 2026) fixes this with a target-affinity token for task-relevant objects and an allocentric grid for global context.
Against Video-3D-LLM:
• SQA3D EM: 58.6 → 62.3
• Scan2Cap CIDEr: 83.8 → 90.5
• Better on all 5 benchmarks tested
Full results across ScanQA, SQA3D, ScanRefer, Multi3DRefer, and Scan2Cap, plus how the central/peripheral split works:
Ask a doctor about vitamin supplements and they will probably tell you that all they do is help you produce “expensive urine”. But the evidence is more positive than that advice might suggest
“Ask the legal team, it’s not in the docs” is apparently something you can automate now.
Deel’s lawyers kept correcting Akai on the same clause. It spotted the pattern and proposed updating the playbook.
The stuff you usually learn after 6 months sitting next to someone.
“Ask Him, Since He Has Been In Politics, What Has He Brought For The Igbos? Look At The Bad State And Erosion In Anambra For a Long Time. Since He Started Politics, What Has He Done About It? Then, When It’s Time For Election, You Come And Start Playing With Our Emotions And Still Want Us To Waste Our Votes On You And Remain Where We Are. It Won’t Happen Again. 2027 Will Be Different. We Igbos Will Get Our Own Share This Time Around By Voting Right.” ~ Anambra State Governor, Chukwuma Soludo, Backs Tinubu And Blasts Peter Obi 👀