🛰️ Daily AI Frontier
‹ back to 2026-07-27

Indexing: the Beginning and the End

Research LLMs & Foundation Models

Merged summary

TL;DR - This paper uses the indexing primitive to expose information bottlenecks across sequence-model architectures. It proves that input ordering, masking, and attention type fundamentally affect whether indexing can be solved with constant depth.

  • Low-causal-complexity masked architectures cannot solve end-position indexing in any constant number of layers.
  • Small softmax transformers solve end-position indexing in one layer; non-masked linear attention requires two.
  • With the index first, small RNNs need one layer, while the other studied architectures need two.
  • The unconditional lower bounds allow infinite-precision arithmetic, and experiments up to sequence length 64 qualitatively support the theory.

Sources (1)

Indexing: the Beginning and the End

arXiv cs.LG Alexander Kozachinskiy, Vicente Opazo, Felipe Urrutia 2026-07-24 arXiv:2607.22361

TL;DR - This paper uses the indexing primitive to expose information bottlenecks across sequence-model architectures. It proves that input ordering, masking, and attention type fundamentally affect whether indexing can be solved with constant depth.

  • Low-causal-complexity masked architectures cannot solve end-position indexing in any constant number of layers.
  • Small softmax transformers solve end-position indexing in one layer; non-masked linear attention requires two.
  • With the index first, small RNNs need one layer, while the other studied architectures need two.
  • The unconditional lower bounds allow infinite-precision arithmetic, and experiments up to sequence length 64 qualitatively support the theory.
item →