I’ve been building TTA @TTAlliance_ for a while now, a community where everyone’s welcome 🤝
✅ Free to join
✅ Free to enter all giveaways
✅ No fees, no gatekeeping
Engage a little & you can even get the VIP role ⭐
NFTs are slowly making their comeback… and we’re here for it.
Recently cook: RAAC — Free mint to 0.11 ETH+
Come hang out, win some stuff, and be part of the journey.
If you’re a CM or Alpha caller, feel free to reach out.
Discord Link: https://t.co/cWFoepFlwo
What does it mean when a 1B model starts landing results in the same range as much larger 7B+ models on reasoning tasks?
Still thinking about the HRM-Text benchmark video.
~40B tokens
~1/350th FLOPS vs larger models
base model only, no post-training
It’s not just “better efficiency” on the chart it sits in a different region entirely, where lower compute and higher scores actually overlap.
@Sapient_Int so is scaling still the main story, or are we already shifting into architecture being the real bottleneck?