Late interaction seems like it could be the sweet spot between full cross-attention and bi-encoders. Curious how it scales with really large document collections in production, and if we really want that in any case vs. optimizing compaction methods 🙂
RESEARCH
-
Researchers Urge OpenAI to Open Source Sora After Shelving
By
–
I am sorry for all the researchers and engineers who worked on Sora & Sora-related projects I hope they don’t just bury it and rather share your hardwork and research with the opensource community to live forever
-
FutureX: Evaluating AI Reasoning and Anticipation in Production
By
–
FutureX tests what most benchmarks miss: Can AI reason, use tools, and anticipate outcomes that haven't happened yet? AI in production isn't about generating content. It's about making decisions in dynamic, uncertain environments. This is where we're seeing strong performance
-
Apollo: Humanoid Robot from Apptronik and NASA Valkyrie
By
–
Apollo: The Humanoid #Robot Built on Apptronik’s Legacy and NASA’s Valkyrie Expertise
— Ronald van Loon (@Ronald_vanLoon) 24 mars 2026
by @Apptronik#Robotics #MachineLearning #ArtificialIntelligence #ML #Innovation pic.twitter.com/8gVtUWzkLTApollo: The Humanoid #Robot Built on Apptronik’s Legacy and NASA’s Valkyrie Expertise
by @Apptronik #Robotics #MachineLearning #ArtificialIntelligence #ML #Innovation -
Brazil’s IBGE Releases Species Richness World Map, Invites Design Critique
By
–
Exactly what you suggest Super strict control We don't NEED to take anyone in that doesn't have the means to support themselves and we NEED to put locals (citizens) first, and not ever give welfare to guests
-
Machine Learning Loss Functions: Comprehensive Guide and Review
By
–
Machine Learning loss functions – infographic Loss Functions in ML Explained: https://
datacamp.com/tutorial/loss-
function-in-machine-learning
… Comprehensive Review of Loss Functions in Deep Learning: https://
arxiv.org/pdf/2504.04242
v1
…
————————
#AI #DataScience #DataScientist #Mathematics -
JEPA Is an Architecture Class, Not a Single Method
By
–
JEPA is a class of architectures, not a method. There were lots of JEPAs before I proposed the name as a unifying concept. For example, Siamese nets (NIPS 1993) are a special case of JEPA (where the predictor is the identity, and the training loss is contrastive as in SimCLR).
-
Multi-Agent Modeling and Geographic Information Systems Practice
By
–
Agent-Based Modeling and Geographical Information Systems: A Practical Primer (GeoSpatial Analytics and GIS) https://amzn.to/3b26CK9 #DataScience #AI #ComputationalScience #SocialScience #NetworkScience #SpatialAnalysis #Simulation
-
ABM Agent-Based Modeling Explained with Resources
By
–
ABM (Agent-Based Modeling) explained: https://
informs-sim.org/wsc09papers/00
9.pdf
…
+—+
And see this additional explanation (source for the attached graphic): https://
bankofengland.co.uk/quarterly-bull
etin/2016/q4/agent-based-models-understanding-the-economy-from-the-bottom-up
… -
Anthropic’s Auto Mode for Claude Code CLI
By
–
Anthropic released Auto Mode for Claude Code CLI, which allows Claude to make its own decisions on which permissions to accept. It is only available on the Team plan in research preview for now. On the desktop app, it is not yet available, but it is in the works.