cs.LGMay 21, 2026

Structured-Sparse Attention for Entity Tracking with Subquadratic Sequence Complexity

Authors: Hangyue ZhaoPaul CaillonErwan FagnouAlexandre Allauzen

Organizations: ESPCI PSL, Paris, France · LAMSADE, Université Paris Dauphine - PSL, Paris, France

Abstract

Entity tracking requires maintaining and updating latent states for entities and attributes over long sequences. Recent task-specific attention operators can compress deep Transformer stacks into a few layers by performing multi-hop state propagation within a single layer, but their dense evaluation remains expensive. We show that in this setting, learned attention is strongly structured: most mass concentrates in local block-diagonal neighborhoods with a light cross-block residue. Exploiting this, we derive a blockwise evaluation of a resolvent-style operator that keeps within-block interactions exact and routes cross-block interactions through a reduced system. The resulting evaluation is subquadratic in sequence length O(n4/3d)O(n^{4/3}d) (and O(n7/3)O(n^{7/3}) when dnd\approx n). On controlled tracking benchmarks, our method matches the dense operator's accuracy while reducing wall-clock time by 1229%12-29\% under a standardized measurement protocol, and is up to 2.4×2.4 \times faster than a compact dense Transformer at comparable exact-match accuracy. We further provide ablations over block size and model capacity, and identify a limitation: performance collapses when the number of simultaneously evolving properties exceeds the number of attention heads.

Explore similar work

CardsList