An embedding approach that loses to SOTA but works in a qualitatively different way
mayfer · x · 2026-10-23
In a thread with @nickcammarata, the author shares an embedding experiment: the method underperforms SOTA embedding search models on benchmarks, but the fact that it works at all — in a qualitatively different way — makes it worth pursuing. Results were from the top 10k English words, his usual first test setup for embedding ideas.
More from Research
- Embedding arithmetic demo: cartoon − animation − comic = diagram — mayfer · 2026-10-23
- Domingos Cites New Result Claiming RL Is 'Basically Useless' vs Supervised Learning — pmddomingos · 2026-10-23
- Wolfram Opens Rule 30 Prizes to AIs: $30,000 for Three Open Problems — cephaloform · 2026-10-23
- Stanford and Databricks use AI agents to reproduce medical findings and test analytical robustness — ChengleiSi · 2026-10-23
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23