💡Text + graphs have powered NLP tasks for a long time. But key questions remain: How do their representations relate to each other during learning? How does that relationship impact task performance? Under what conditions is integrating text and graph more beneficial? In our paper, "R²-CoD: Understanding Text-Graph Complementarity in Relational Reasoning via Knowledge Co-Distillation", we take an analysis-driven perspective and uncover a spectrum of behaviors from complementarity → partial alignment → full alignment. 📜arxiv.org/abs/2508.01475 🧵[1/8]
MACHINE LEARNING
-
Scobleizer builds Huggingface robot, trains policies, asks Grok if it’s a policy
By
–
Hmm, I'm building my @huggingface robot, which looks like a fun toy. But will train it with a bunch of policies. I see others who have already built theirs and they put @Grok inside. Hey @Grok are you a policy?
-
Open-source topic modeling and data pipeline project
By
–
Si vous voulez faire le projet de topic modeling votre côté, vous pouvez repartir de mon Git : il y a les données dessus et un dashboard que j'ai dev pour en récupérer davantage, avec un squelette pour accueillir des pipelines. D'ici quelques semaines, j'y mettrai une version x.com/DFintelligence…
-
Open-source topic modeling pipeline and dashboard project
By
–
Si vous voulez faire le projet de topic modeling votre côté, vous pouvez repartir de mon Git : il y a les données dessus et un dashboard que j'ai dev pour en récupérer davantage, avec un squelette pour accueillir des pipelines. D'ici quelques semaines, j'y mettrai une version
-
Design AI to drive without connection: exactly what Tesla did
By
–
The right way is to design the AI to drive without any connection at all. Oh, wait, that's exactly what Tesla did.
-
GLM-4.7 Foundation Model Achieves Record-Breaking Reasoning and Coding Scores
By
–
Now on Poe: GLM-4.7 This next-generation foundation model, powered by @novita_labs on Poe, posts record‑breaking scores across several major reasoning and coding benchmarks. (1/2)
-
All Learning Happens Online Without Pretraining or Transfer
By
–
It’s ALL online learning in nature — no pretraining and transfer.
-
AI will read X and find funniest memes in 2026
By
–
Oh, great, AI will read X and post more memes in 2026. Will need to have an AI read those memes and tell me which ones are the funniest. 🙂
-
New fast-learning AIs: World Models trend in 2026, Tesla leads
By
–
Yup. And new AIs are on the way that learn faster. Which is why World Models will be a huge trend in 2026. Who has the best World Model? Tesla.
-
Centralization for routing, Tesla ensures safe offline operation
By
–
Yes. Now there is a role for centralization. Where should the robot drive to pick up next rider, for instance. But because of how Tesla was trained it will be able to complete the job even if everything goes down. And do it safely.