arXiv subjects

Tomer Schlank

Publications and source records attributed to Tomer Schlank.

9 recordsLinked to original sources

arXiv2025

Higher Zariski Geometry

We revisit the classical constructions of tensor-triangular geometry in the setting of stably symmetric monoidal idempotent-complete $\infty$-categories, henceforth referred to as 2-rings. In this setting, we produce a Zariski topology, a Zariski spectrum, a category of locally 2-ringed spaces (more generally $\infty$-topoi), and an affine spectrum-global sections adjunction, based on the framework of ``$\infty$-topoi with geometric structure'' as developed by Lurie in \cite{LurieDAG5}. Using work of Kock and Pitsch, we compute that the underlying space of the Zariski spectrum of a 2-ring recovers the Balmer spectrum of its homotopy category. These constructions mirror the analogous structures in the classical Zariski geometry of commutative rings (and commutative ring spectra), and we also demonstrate additional compatibility between classical Zariski and higher Zariski geometry. For rigid 2-rings, we show that the descent results of Balmer and Favi admit coherent enhancements. As a corollary, we obtain that the Zariski spectrum fully faithfully embeds rigid 2-rings into locally 2-ringed $\infty$-topoi. In an appendix, we prove a ``stalk-locality principle'' for the telescope conjecture in the rigid setting, extending earlier work of Hrbek.

math.AG↗

arXiv2024

Looking Beyond The Top-1: Transformers Determine Top Tokens In Order

Understanding the inner workings of Transformers is crucial for achieving more accurate and efficient predictions. In this work, we analyze the computation performed by Transformers in the layers after the top-1 prediction has become fixed, which has been previously referred to as the "saturation event". We expand the concept of saturation events for top-k tokens, demonstrating that similar saturation events occur across language, vision, and speech models. We find that these saturation events happen in order of the corresponding tokens' ranking, i.e., the model first decides on the top ranking token, then the second highest ranking token, and so on. This phenomenon seems intrinsic to the Transformer architecture, occurring across different architectural variants (decoder-only, encoder-only, and to a lesser extent full-Transformer), and even in untrained Transformers. We propose an underlying mechanism of task transition for this sequential saturation, where task k corresponds to predicting the k-th most probable token, and the saturation events are in fact discrete transitions between the tasks. In support of this we show that it is possible to predict the current task from hidden layer embedding. Furthermore, using an intervention method we demonstrate that we can cause the model to switch from one task to the next. Finally, leveraging our findings, we introduce a novel token-level early-exit strategy, which surpasses existing methods in balancing performance and efficiency.

cs.CL↗

arXiv2021

Rational motivic path spaces and Kim's relative unipotent section conjecture

We initiate a study of path spaces in the nascent context of "motivic dga's", under development in doctoral work by Gabriella Guzman. This enables us to reconstruct the unipotent fundamental group of a pointed scheme from the associated augmented motivic dga, and provides us with a factorization of Kim's relative unipotent section conjecture into several smaller conjectures with a homotopical flavor. Based on a conversation with Joseph Ayoub, we prove that the path spaces of the punctured projective line over a number field are concentrated in degree zero with respect to Levine's t-structure for mixed Tate motives. This constitutes a step in the direction of Kim's conjecture.

math.AG↗

arXiv2020

Brauer and Etale Homotopy Obstructions to Rational Points on Open Covers

In 2010, Poonen gave the first example of failure of the local-global principle that cannot be explained by Skorobogatov's étale Brauer-Manin obstruction. Motivated by this example, we show that the Brauer-Manin obstruction detects non-existence of rational points on a sufficiently fine Zariski open cover of any variety over an imaginary quadratic or totally real field. We provide some evidence for why this is expected to happen more generally over any number field, some of which relates to the section conjecture in anabelian geometry. We then prove a result about the behavior of the étale Brauer obstruction in fibrations of varieties using the étale homotopy obstruction of Harpaz and the second author. We finally use that result and other techniques to further analyze Poonen's example in light of our general results.

math.NT↗

arXiv2020

Chromatic homotopy theory is asymptotically algebraic

Inspired by the Ax--Kochen isomorphism theorem, we develop a notion of categorical ultraproducts to capture the generic behavior of an infinite collection of mathematical objects. We employ this theory to give an asymptotic solution to the approximation problem in chromatic homotopy theory. More precisely, we show that the ultraproduct of the $E(n,p)$-local categories over any non-prinicipal ultrafilter on the set of prime numbers is equivalent to the ultraproduct of certain algebraic categories introduced by Franke. This shows that chromatic homotopy theory at a fixed height is asymptotically algebraic.

math.AT↗

arXiv2019

Morphisms of rational motivic homotopy types

We investigate several interrelated foundational questions pertaining to the study of motivic dga's of Dan-Cohen--Schlank [8] and Iwanari [13]. In particular, we note that morphisms of motivic dga's can reasonably be thought of as a nonabelian analog of motivic cohomology. Just as abelian motivic cohomology is a homotopy group of a spectrum coming from K-theory, the space of morphisms of motivic dga's is a certain limit of such spectra; we give an explicit formula for this limit --- a possible first step towards explicit computations or dimension bounds. We also consider commutative comonoids in Chow motives, which we call ``motivic Chow coalgebras''. We discuss the relationship between motivic Chow coalgebras and motivic dga's of smooth proper schemes. As a small first application of our results, we show that among schemes which are finite étale over a number field, morphisms of associated motivic dga's are no different than morphisms of schemes. This may be regarded as a small consequence of a plausible generalization of Kim's relative unipotent section conjecture, hence as an ounce of evidence for the latter.

math.AG↗

arXiv2019

The $\infty$-Categorical Eckmann-Hilton Argument

We define a reduced $\infty$-operad $\mathcal{P}$ to be $d$-connected if the spaces $\mathcal{P}\left(n\right)$, of $n$-ary operations, are $d$-connected for all $n\ge0$. Let $\mathcal{P}$ and $\mathcal{Q}$ be two reduced $\infty$-operads. We prove that if $\mathcal{P}$ is $d_{1}$-connected and $\mathcal{Q}$ is $d_{2}$-connected, then their Boardman-Vogt tensor product $\mathcal{P}\otimes\mathcal{Q}$ is $\left(d_{1}+d_{2}+2\right)$-connected. We consider this to be a natural $\infty$-categorical generalization of the classical Eckmann-Hilton argument.

math.AT↗

arXiv2017

Normal amenable subgroups of the automorphism group of the full shift

We show that every normal amenable subgroup of the automorphism group of the full shift is contained in its center. This follows from the analysis of this group's Furstenberg topological boundary, through the construction of a minimal and strongly proximal action. We extend this result to higher dimensional full shifts. This also provides a new proof of Ryan's Theorem and of the fact that these groups contain free groups.

math.DS↗

arXiv2011

A Note on the Entropy/Influence Conjecture

The entropy/influence conjecture, raised by Friedgut and Kalai in 1996, seeks to relate two different measures of concentration of the Fourier coefficients of a Boolean function. Roughly saying, it claims that if the Fourier spectrum is "smeared out", then the Fourier coefficients are concentrated on "high" levels. In this note we generalize the conjecture to biased product measures on the discrete cube, and prove a variant of the conjecture for functions with an extremely low Fourier weight on the "high" levels.

math.CO↗