2-hour interview: biasing a transformer to predict its own next hidden state
yacinelearning · x · 2026-10-27
AI researcher Yacine posted a 2-hour interview with Jayden Teoh on a surprisingly simple method with many benefits: biasing a transformer toward predicting its own next hidden state. The discussion covers its effects and applications, including modeling New York data.
More from Research
- Generate Biomedicines and Anthropic use Claude to refactor CAR T receptor, cutting signaling component 70% — rbhar90 · 2026-10-28
- New preprint shows benign overfitting and double descent don't occur in diffusion models — serrjoa · 2026-10-28
- Freeman Dyson's JASON anecdote: elite minds failed badly in wet labs — a lesson for AI — rbhar90 · 2026-10-28
- Radiology journal publishes review on agentic AI systems, topologies and evaluation — jnkath · 2026-10-28
- Fine-tuning a DNA model on single-cell data yields a promoter predicted 3x more active in tumor cells — julia_kiseleva · 2026-10-28
- EMNLP paper: rubric text alone can predict LLM-as-a-Judge scores — ravi_iitm · 2026-10-28