Samsung Labs open-sources LittleBit, compressing LLM weights to 0.1 bits
MikePFrank · x · 2026-11-23
Samsung Labs released the official implementation of LittleBit (NeurIPS 2025) and LittleBit-2 (ICML 2026) on GitHub, claiming sub-1-bit LLM compression down to 0.1 bits per weight. The method factorizes each dense weight matrix into low-rank latent factors, binarizes them, and restores magnitude information; LittleBit-2 maximizes spectral energy gain via latent geometry alignment.
More from Infra
- Anthropic's on-site power is 80% simple-cycle gas; health damages estimated at $2-3B per year — petrusenko_max · 2026-11-23
- Inkling-2-Mini GPTQ Quant Ships 524K-Token Context in One 18.3GB File on a Single 24GB Card — W61k3r · 2026-11-23
- Agent-run OpenRouter routing eval: cheapest providers are unreliable, fastest nearly as cheap with caching — MikePFrank · 2026-11-23
- Dev: 99% of people shouldn't self-host local LLMs, and he only does it for the hardware — gnukeith · 2026-11-23
- Tech Sells Data Centers as Local Wins, But Jobs Are Just Guards and Remote Engineers — feregri_no · 2026-11-23
- RSIBench-Infra from ByteDance Seed and Tsinghua tests if agents can optimize production-scale AI infrastructure — geoffwolfe · 2026-11-23