Meta's Brain2Qwerty v2 Hits 61% Word Accuracy
petetheposter · reddit · 2026-07-07
Meta unveiled Brain2Qwerty v2, a non-invasive brain-to-text pipeline trained on roughly 22,000 sentences typed by 9 subjects under MEG. It achieved an overall word accuracy of 61%, with the best subject reaching 78%. Meta simultaneously open-sourced the training code for v1 and v2, and BCBL released the v1 dataset. The significance lies in the fact that non-invasive BCI has long lagged behind invasive solutions; if these results hold, it represents substantial progress. Furthermore, open-sourcing the code and data ensures its value extends beyond a mere headline or benchmark screenshot.
Related event: Meta Unveils Brain2Qwerty Non-Invasive Brain-to-Text AI(2 posts)→
More from Research
- Baseten study finds new facts in LLM weights are fragile unless trained from many restatements — alex_verem · 2026-07-21
- Kimi K3 and Fable 5 now look much closer than the old open-vs-closed gap — FinanceYF5 · 2026-07-21
- uv-scripts/ocr returns to the top of Hugging Face datasets with a JSON model picker — vanstriendaniel · 2026-07-21
- DeepSearch-World trains web agents with 420K verifiable QA tasks — HKUST · 2026-07-21
- GigaAM Multilingual targets low-resource Central Asian ASR with 2M hours of audio — ai-sage · 2026-07-21
- WorldCupArena benchmarks language models on 104 football matches — Zhaokai Wang · 2026-07-21