HOT RADAR
llmHEAT 13.1°

AI CURATED HOT STORY · 11/18/2026

Latent Lattice Language Models Cut KV Cache 45% with 22 Layers

1 reports archived1 independent sourcesupdated 11/18/2026, 06:17:18
Synthesis & Latest Updates
1 Sources Cross-Validated

The paper introduces Latent Lattice Language Models (LLLM), which add a horizontal latent width alongside the vertical residual-stream width. At 20B-token pretraining scale, LLLM matches a 40-layer Transformer's validation perplexity using only 22 layers, cutting KV cache by 45%, and can be combined with a Jacobi refiner for further gains.

LATEST/The paper introduces Latent Lattice Language Models (LLLM), which add a horizontal latent width alongside the vertical residual-stream width. At 20B-token pretraining scale, LLLM matches a 40-layer Transformer's validation perplexity using only 22 layers, cutting KV cache by 45%, and can be combined with a Jacobi refiner for further gains.

Hourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first