In the first part of this talk, we will discuss a single-cut rule for late-time cosmological correlators, which constitute an important class of observables in de Sitter cosmology. We will derive this rule directly from perturbative unitarity and the Hermitian analyticity of the propagators. A key point of the analysis is that, after cutting a correlator, the resulting expression cannot in general be written solely in terms of lower-point correlators. Our formalism therefore introduces a new object that is absent from previously available cutting formulae in the literature. Our perspective is to use the cutting rule to reduce an exchange-level correlator to a combination of contact-level data, involving correlators and/or wavefunction coefficients, which are technically simpler to compute. In the second part of the talk, we will show that once the cut of an exchange-level correlator is known in terms of such contact-level objects, this information can be used to reconstruct the full correlator. The reconstruction is based on dispersion relations, familiar from complex analysis. We will conclude by showing that the same dispersion-relation framework naturally leads to a set of diagrammatic rules for computing arbitrary tree-level cosmological correlators.
Communication complexity studies how much information must be exchanged to solve a problem whose input is split among two or more parties. The classical setting deals with Boolean inputs split between two parties. In the algebraic variant, inputs, messages, and
computation are over some (potentially infinite) field. This variant, introduced by Abelson in 1978 and developed by Grigoriev in 2008, is relatively unexplored. In this talk, I will describe this setting, the hurdles to directly applying Boolean techniques, and the techniques that do work to give lower bounds. I will describe some specific lower bounds we establish. More details can be found in the preprints
https://arxiv.org/abs/2609.17082 and
https://eccc.weizmann.ac.il/report/2026/182/
This is joint work with Manon Blanc, Prateek Dwivedi, Magnus Hansen, and Nutan Limaye, at ITU Copenhagen.
This talk will be in hybrid mode. Please find zoom link below:
https://zoom.us/j/99598370034
Meeting ID: 995 9837 0034
Passcode: 941422
Long-range correlations are used to probe collective behavior in many-body systems. In finite systems, however, global conservation of
energy, momentum, particle number, and other conserved charges can itself generate correlations between otherwise independent degrees of
freedom. It is therefore crucial to remove the contribution of these conservation-induced correlations from observed long-range correlations for
understanding the nature of the emergent collectivity.
In this talk, I will present a unified conditional-probability framework in which equilibrium statistics and conservation-induced correlations
emerge from conditioning on additive conserved quantities. The one-mode conditional law yields the Maxwell–Boltzmann, Bose–Einstein, and
Fermi–Dirac distributions at leading saddle order, while the two-mode law gives the leading finite-rank covariance induced by exact
conservation. This structure allows us to construct observables orthogonal to selected conserved quantities, for which the leading
conservation-induced covariance vanishes. As an application, I will discuss long-range correlations in high-energy p-Pb collisions, which
exhibit patterns resembling collective behaviour in collisions of large nuclei where they are commonly associated with hydrodynamic collective
response. Conservation effects are particularly relevant in these smaller systems because fewer particles are produced. Using PYTHIA generated
events, I will illustrate how conservation-aligned and conservation-orthogonal contributions can be separated, and discuss the implications for
interpreting the emergent collective behaviour. [Based on arXiv:2606.29050]
Ideological reversal occurs in an election when parties choose policy platforms that reverse their ideological order. We study an ordinal spatial voting game in which voter and party preferences are single peaked. When every voter participates in voting, single peakedness alone prevents reversal at equilibrium. However, when voters are allowed to abstain, we show that single peakedness is insufficient to rule out such reversals. We demonstrate an example where parties are ideologically reversed at the unique equilibrium. We then identify a sufficient condition on party preferences which rules out the existence of such reversed equilibria, even under abstention.
Real numbers may have infinitely many digits yet admit short, finite descriptions. Can we recover such a description of an unknown number using only comparisons? This question extends classical integer/rational search to richer representations and connects numerical search with exact query learning.
This talk will focus on automatic reals, whose digits are determined by finite automata. For an unknown real with an s-state representation, we obtain a polynomial-time algorithm using O(s^2) comparisons. For this algorithm, we revisit Angluin’s exact learning model, well known in learning theory, and show how automata-learning techniques can be adapted to identify unknown real numbers using only comparison queries.
If time permits, we will discuss other representations, including algebraic numbers.
This is ongoing joint work with Arvind.
This talk will be on certain new techniques for triangulated categories introduced by Neeman in a series of papers, which have been used to prove various new results and settling multiple open conjectures. We will give a broad overview of these techniques, and illustrate various applications in algebraic geometry. In particular, we will see a representability theorem and its application to a new proof of GAGA theorems, a result on transferring semiorthogonal decompositions, and a result on the non-existence of bounded t-structures for perfect complexes on singular finite-dimensional Noetherian schemes.