AToM project open-sources retrofitted Qwen3-8B-DMS-8x and Bolmo-7B with NVIDIA and AI2
PontiEdoardo · x · 2026-09-11
In his AToM recruiting thread, Edoardo Ponti reveals the team has partnered with NVIDIA and Allen AI to release retrofitted open-weight models: Qwen3-8B-DMS-8x and Bolmo-7B, porting adaptive tokenization/memory methods onto existing frontier open models, with plans to keep contributing to the open-weight community. Resources include project-owned B200s, cloud credits, H200s from Imperial, and funding for 2 conferences per year.
Related event: AToM Open-Sources Adapted Qwen3 and Bolmo Models with NVIDIA and AI2(2 posts)→
More from Research
- ParetoRSI-Gym benchmarks whether AI agents can recursively self-improve LLMs across domains — Haojin Wang · 2026-10-28
- Stanford's Ψ-0.6 is a unified vision-language world model with adaptive reasoning streams — Stanford · 2026-10-28
- Shanghai AI Lab's OpenESM shows energy-steered models can scale as foundation learners — Shanghai-AI-Laboratory · 2026-10-28
- CellTrack-Bench tests VLMs on microscopic cell tracking, exposing reasoning gaps — biohub · 2026-10-28
- Hybrid LLM agent scores 83.2%, beats nine VLA policies on new SO-101 robot arm benchmark — Truman Hickok · 2026-10-28
- DeepMind researcher: all of mathematics will soon be formalizable, with agents hunting proofs — ChrSzegedy · 2026-10-28