Simplex launches research blog Belief Updates, showing transformer belief geometry forms telescoping cones
rishabh16_ · x · 2026-09-10
AI startup Simplex has launched its research blog Belief Updates with two inaugural posts.
- A welcome post explaining the blog's motivation.
- New research on LLM internals: since training data mixes many sources, the team asks what structure to expect in activations. They show the belief geometry over such data forms telescoping cones, and that transformers represent this structure.
More from Research
- DeepMind podcast: AI weather models now aid Hurricane Melissa forecasts — PeterWBattaglia · 2026-09-10
- Giving an agent a shell widens the context-format gap from 50 to 74 points — arch1v1sor · 2026-09-10
- Project CETI's AI uncovers vowel-like patterns in sperm whale vocalizations — begusgasper · 2026-09-10
- Geodesic Launches AI-Native Protein Therapeutics Platform With NovaDDE, NovaAtom-Lite — QuanquanGu · 2026-09-10
- OpenAI's Navier-Stokes result used 130B tokens — Reddit debate on whether it's just brute force — kaljakin · 2026-09-10
- Autoresearch Loop with Tinker Reproduces Self-Distillation Papers at Predictable Cost — SRSchmidgall · 2026-09-10