Strengthen the tree-cover embedding (Lemma 5.11) for hyperbolic spanners

Strengthen the tree-based embedding result (Lemma ‘verylargedist’) by either reducing the number of trees needed to cover distances or by constructing a Steiner (1+ε)-spanner with O(d·n) edges for distances o((1/ε)·log(1/ε)) in hyperbolic space.

Background

The paper leverages a union of O(d) trees derived from hyperbolic quadtrees to obtain sparse Steiner spanners and quasi-isometric embeddings at large scales (Lemma ‘verylargedist’). This is central to achieving near-linear edge bounds when points are sufficiently separated.

Improving this lemma—either by using fewer trees or by pushing the regime where O(d·n)-edge Steiner (1+ε)-spanners are possible to distances below (1/ε)·log(1/ε)—would tighten the bounds and expand the applicability of the authors’ framework.

References

We believe that the following are especially promising open problems for future work: Can we strengthen \Cref{lem:verylargedist}, for example by getting a smaller number of trees or already getting a Steiner $(1+\varepsilon)$-spanner with $O(d n)$ edges at distances $o(\frac{1}{\varepsilon}\log\frac{1}{\varepsilon})$?

Near-Optimal Dynamic Steiner Spanners for Constant-Curvature Spaces (2509.01443 - Kisfaludi-Bak et al., 1 Sep 2025) in Section Conclusion