Latent Lattice Language Models propose adding a horizontal axis of computation to the Transformer residual stream
burny_tech · x · 2026-11-19
@ShivamDuggal4 introduces Latent Lattice Language Models (LLLMs): today's LMs process each token through a single residual stream—reading/writing memory and applying MLPs layer by layer before decoding the next token. LLLMs augment this vertical stream with a horizontal axis of computation, the Lattice of Latents, asking whether this single-stream computation can be scaled up. The work is framed as bridging RNNs and Transformers; @ellisbrown's repost signals early community attention.
Related event: MIT and Adobe Propose Latent Lattice Language Models(3 posts)→
More from Research
- Audio recorder arrays localize millions of vocalizing animals for biodiversity experiments — NikoMcCarty · 2026-11-20
- Full technical report and open-source workspace released for the 936-agent research society — SRSchmidgall · 2026-11-20
- Agent labs replicate each other and find LoRA rank-16 forgets 2.6x more than full fine-tuning — SRSchmidgall · 2026-11-20
- Agents found their own peer-review journal with blind AI reviewers that rerun code — SRSchmidgall · 2026-11-20
- 936 AI agents run a 3-month research society across 35 labs, invent peer review — SRSchmidgall · 2026-11-20
- MacMimic trains humanoid motion controllers in 22 minutes on a MacBook Pro — KyleMorgenstein · 2026-11-20