SOURCE-LINKED INTELLIGENCE
Latest Exact Match Attention
We introduce latest exact match attention (LEMA), an attention variant for transformers where queries and keys are binarized and each query attends only to the latest exactly matching key. We prove that LEMA transformers with chain of thought can simulate word-RAMs, as was recently shown for the less restrictive rightmost hard attention. In contrast to prior hard attention variants, the restriction to exact matches enables an efficient converse direction: word-RAMs can simulate LEMA transformers at a cost per token independent of the context length. Together, these results yield a close corres
Read original source ↗ Open in workspace
- recordType
- paper
- region
- Global
Evidence & attribution
- arXiv · AI, language, vision and robotics · 2026-09-22T07:36:00.000Z
First collected: 2026-09-23T04:21:13.910Z. This is not the publication date.