AI2's byte-level LLM retrofitting paper accepted at Nature, Bolmo method extended to Qwen and Llama
allen_ai · x · 2026-10-08
The Allen Institute for AI announced that its paper on retrofitting language models to operate directly over bytes — the approach behind Bolmo, using latent tokenisation — has been accepted by Nature.
The team is also releasing new checkpoints that extend the method from Olmo to Qwen and Llama, bringing byte-level latent tokenisation to these popular open model families. Details in the linked paper.
More from Models
- User stunned as Grok spots legal file errors Claude and ChatGPT missed for months — iruletheworldmo · 2026-10-08
- The lab that doesn't nerf its flagship model two weeks after release will win — StewartalsopIII · 2026-10-08
- Ask: Is Qwen 3.8 Flash-Next Worth It Over 3.6 35B-A3B on 64GB RAM + 16GB VRAM? — msalsas · 2026-10-08
- Blind Test Pits AI Models Against Real Community and Social Media Tasks — thisiskp_ · 2026-10-08
- theo challenges Sentry CEO: same AI coding tasks now over 10x cheaper than a year ago — theo · 2026-10-08
- theo counters Zeeg: same task 55x cheaper in a year with Gemini 3 Flash vs o1 — theo · 2026-10-08