BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

John Sous

John Sous contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
cond-mat.str-elcond-mat.quant-gascond-mat.mes-hallMachine Learningphysics.atom-phquant-phArtificial Intelligencecond-mat.mtrl-scihep-thphysics.chem-ph

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
8works
0followers
10topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

8 published item(s)

preprint2026arXiv

Memory Inception: Latent-Space KV Cache Manipulation for Steering LLMs

Steering large language models (LLMs) is usually done by either instruction prompting or activation steering. Prompting often gives strong control, but caches guidance tokens at every layer and can clutter long interactions; activation steering is compact but typically weaker and does not support large structured reminders. We introduce memory inception (MI), a training-free method that steers in latent attention space by inserting text-derived key-value (KV) banks only at selected layers. Rather than materializing reminder content throughout the prompt cache, MI treats steering as selective KV allocation, injecting latent slots only where the model routes to them. On matched personality-steering tasks, MI gives the best overall control--drift trade-off, remaining competitive with prompting while consistently outperforming CAA. On updateable guidance, MI supports mid-conversation behavior shifts without rewriting the visible transcript, achieving the highest post-shift alignment on Qwen3. On structured reasoning, MI outperforms visible prompting on HARDMath and PHYSICS (10/12 subject$\times$mode cells), serving as proxies for structured reasoning in verifiable domains, while cutting content-matched KV storage by up to 118$\times$. These results position MI as a powerful steering method when guidance is persistent, structured, or expensive to keep in the visible transcript.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Spectral Lens: Activation and Gradient Spectra as Diagnostics of LLM Optimization

Training loss and throughput can hide distinct internal representation in language-model training. To examine these hidden mechanics, we use spectral measurements as practical and operational diagnostics. Using a controlled family of decoder-only models adapted from the modded NanoGPT codebase, we introduce an empirical protocol based on activation covariance and per-sample gradient SVD spectra. This dual-view reveals three empirical findings and one mechanistic explanation. First, batch size acts as a latent determinant of representation geometry: runs that reach equal loss settle into systematically distinct activation spectra. Second, the activation covariance tail measured early in training reliably forecasts downstream token efficiency. Third, movement of the activation spectrum head (leading modes), together with gradient spectra, characterizes underlying learning-dynamics changes, separating learning-side architectural improvements from primarily execution-side gains. These predictive and diagnostic signals persist across the 12-, 36-, and 48-layer model tiers. Finally, a mechanistic model proves the main observations and explains how activation covariance spectra correlate with task-aligned feature learning.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

The normal state of attractive Fermi gases from coupled-cluster theory

We introduce coupled-cluster (CC) theory for the numerical study of the normal state of two-component, dilute Fermi gases with attractive, short-range interactions at zero temperature. We focus on CC theory with double excitations (CCD) and discuss its close relationship with -- and improvement upon -- the t-matrix approximation, i.e., the resummation of ladder diagrams via a random-phase approximation. We further discuss its relationship with Chevy's variational wavefunction ansatz for the Fermi polaron and argue that CCD is its natural extension to nonzero minority species concentrations. Studying normal state energetics for a range of interaction strengths below and above unitarity, we find that CCD yields good agreement with fixed-node diffusion Monte Carlo. We find that CCD does not converge for small polarizations and large interaction strengths, which we speculatively attribute to the nascent instability to a superfluid state.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Exploring universal and non-universal regimes of trimers from three-body interactions in one-dimensional lattices

We investigate the formation of trimers in an infinite one-dimensional lattice model of hard-core particles with single-particle hopping $t$ and and nearest-neighbour two-body $U$ and three-body $V$ interactions of relevance to Rydberg atoms and polar molecules. For sufficiently attractive $U\leq-2t$ and positive $V>0$ a large trimer is stabilized, which persists as $V\rightarrow \infty$, while both attractive $U\leq0$ and $V\leq0$ bind a small trimer. The excited state above this small trimer is also bound and has a large extent; its behavior as $V\rightarrow -\infty$ resembles that of the large ground-state trimer. These large bound states appear to admit a continuum description. Furthermore, we find that in the limit $V>>t$, $U<-2t$ the bound-state behavior qualitatively evolves with larger $|U|$ from a state described by the scattering of three far separated particles to a state of a compact dimer scattering with a single particle.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Fractons from polarons

Fractons are a type of emergent quasiparticle which cannot move freely in isolation, but can easily move in bound pairs. Similar phenomenology is found in boson-affected hopping models, encountered in the study of polaron systems and hole-doped Ising antiferromagnets, in which motion of a particle requires the creation or absorption of bosonic excitations. We show that boson-affected hopping models can provide a natural realization of fractons, either approximately or exactly, depending on the details of the system. We first consider a generic one-dimensional boson-affected hopping model, in which we show that single particles move only at sixth order in perturbation theory, while motion of bound states occurs at second order, allowing for a broad parameter regime exhibiting approximate fracton phenomenology. We explicitly map the model onto a fracton Hamiltonian featuring conservation of dipole moment via integrating out the mediating bosons. We then consider a special type of boson-affected hopping models with mutual hard-core repulsion between particles and bosons, accessible in hole-doped mixed-dimensional Ising antiferromagnets, in which the hole motion is one dimensional in an otherwise two-dimensional antiferromagnetic background. We show that this system, which is within the current reach of ultracold-atom experiments, exhibits perfect fracton behavior to all orders in perturbation theory, thereby enabling the experimental study of dipole-conserving field theories. We further discuss diagnostic signatures of fractonic behavior in these systems. In studying these models, we also identify simple effective one-dimensional microscopic Hamiltonians featuring perfect fractonic behavior, paving the way to future studies on fracton physics in lower dimensions.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Strongly correlated ladders in K-doped $p$-terphenyl crystals

Potassium-doped terphenyl has recently attracted attention as a potential host for high-transition-temperature superconductivity. Here, we elucidate the many-body electronic structure of recently synthesized potassium-doped terphenyl crystals. We show that this system may be understood as a set of weakly coupled one-dimensional ladders. Depending on the strength of the inter-ladder coupling the system may exhibit spin-gapped valence-bond solid or antiferromagnetic phases, both of which upon hole doping may give rise to superconductivity. This terphenyl-based ladder material serves as a new platform for investigating the fate of ladder phases in presence of three-dimensional coupling as well as for novel superconductivity.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Fractons from frustration in hole-doped antiferromagnets

Recent theoretical research on tensor gauge theories led to the discovery of an exotic type of quasiparticles, dubbed fractons, that obey both charge and dipole conservation. Here we describe physical implementation of dipole conservation laws in realistic systems. We show that fractons find a natural realization in hole-doped antiferromagnets. There, individual holes are largely immobile, while dipolar hole pairs move with ease. First, we demonstrate a broad parametric regime of fracton behavior in hole-doped two-dimensional Ising antiferromagnets viable through five orders in perturbation theory. We then specialize to the case of holes confined to one dimension in an otherwise two-dimensional antiferromagnetic background, which can be realized via the application of external fields in experiments, and prove ideal fracton behavior. We explicitly map the model onto a fracton Hamiltonian featuring conservation of dipole moment. Manifestations of fractonicity in these systems include gravitational clustering of holes. We also discuss diagnostics of fracton behavior, which we argue is borne out in existing experimental results.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Rydberg impurity in a Fermi gas: Quantum statistics and rotational blockade

We consider the quench of an atomic impurity via a single Rydberg excitation in a degenerate Fermi gas. The Rydberg interaction with the background gas particles induces an ultralong-range potential that binds particles to form dimers, trimers, tetramers, etc. Such oligomeric molecules were recently observed in atomic Bose-Einstein condensates. In this work, we demonstrate with a functional determinant approach that quantum statistics and fluctuations have observable spectral consequences. We show that the occupation of molecular states is predicated on the Fermi statistics, which suppresses molecular formation in an emergent molecular shell structure. At large gas densities this leads to spectral narrowing, which can serve as a probe of the quantum gas thermodynamic properties.

0 citations0 reviewsNo code linkedOpen access