Generic Triple-Latent Compression with Gated Associative Retrieval 事件

Generic Triple-Latent Compression with Gated Associative Retrieval arXiv:2606.05175v1 Announce Type: new Abstract: We study generic triple-latent sequence models that maintain a running token state and compressed pair-memory pathway to capture higher-order token interactions without benchmark-specific parsing. The triple-latent family improves a small Transformer baseline on byte-level WikiText-2 and on a tokenizer-based MiniMind language-model benchmark, while a recall-focused gated key-value