SOURCE-LINKED INTELLIGENCE

Shared Global KV with Layer-Specific Local History

arXiv · AI, language, vision and robotics · article · Sep 23, 2026 · UTC

Decoder-only Transformer language models cache keys and values (KV) to reuse past computation during generation. Sharing KV across layers saves storage but reduces the diversity of representations available across depth. We study what local memory should retain alongside shared global KV, separating historical content from the input source used to form it. At 126M parameters and 2K context, an eight-seed study finds about 1.4% lower held-out test perplexity with local history than with a current-token local branch. Capacity, entry-count and training-compute controls support the value of histor

Read original source ↗ Open in workspace

recordType

paper

region

Global