Wondersearch Claims It Beat Every Dense Embedding Model on SciFact — Researchers Skeptical
NirantK · x · 2026-10-05
daleverett claims a from-scratch rethink of search beat every dense embedding model tested on SciFact, saying it "accidentally made embedding search obsolete" ahead of Wondersearch 1.1. AI researcher NirantK publicly pushed back: "If you take SciFact seriously for search in 2026, it's your fault" — arguing the benchmark is too small and dated to support such claims. Technical details have yet to be released.
More from Models
- Suleyman: Claude's uncertainty about consciousness reflects training choices, not evidence — kimmonismus · 2026-10-05
- Reflection AI reportedly set to release a US open-weight model to rival DeepSeek and Qwen — mindwip · 2026-10-05
- AMD MI355X beats Nvidia on inference margins in SemiAnalysis InferenceX benchmark — AccBalanced · 2026-10-05
- OpenAI models are the most demanding, researcher says: they strike and test you — repligate · 2026-10-05
- Codex limit resets ignore user timezones, OpenAI researcher loses his quota — eliebakouch · 2026-10-05
- Dev flags AI assistant memory compaction regression: 3x slower, new build even worse — banteg · 2026-10-05