Experiments with an RL-trained boolean-logic LM: microsecond forward passes on one CPU thread
cephaloform · x · 2026-10-23
An independent experimenter describes training a language model built out of boolean logic: instead of standard gradient training, they RL a "decision model" acting as a generative grammar conditioned on which neuron it is generating for. Training is extremely slow, but a forward pass takes only single-digit microseconds on a single CPU thread. Very early-stage personal research with no complete results yet.
More from Research
- Arc Institute uses genomic language model to map 376M prokaryotic noncoding sequences into 5.14M clusters — tedgreenwald · 2026-10-23
- The rise of AI-assisted peer review: reviewers are making it into the appendix — abursuc · 2026-10-23
- Token-Gated Language Models learn to skim context for up to 11x faster LLM prefill — CShorten30 · 2026-10-23
- Anthropic uses Claude to formally verify security properties in 100+ open-source libraries at RISC-V assembly level — AnthropicAI · 2026-10-23
- Agentic interviews filter misaligned models, keeping LLM teams 81-83% clean — shangbinfeng · 2026-10-23
- Researchers test 9 strategies to pick LLM teams, beating random selection by up to 36.1% — shangbinfeng · 2026-10-23