Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Friday 6 November

High Energy Physics (Theory)

Holographic warped CFTs on AdS$_3$, parity violation and unitarity

We consider recently introduced boundary conditions for General Relativity on AdS$_3$, whose canonical generators span the warped conformal algebra, in the case of the Einstein-Hilbert action supplemented by parity-odd terms that do not modify the field equations in vacuum. The canonical generators are shown to acquire nonlinear contributions, and their algebra remains being the semidirect sum of Virasoro and $u(1)$ Kac-Moody, but with shifted central extensions. Thus, for suitable values of the parity-odd coupling, the Kac-Moody level can be nonnegative ($k_\text{KM}\geq 0$), admitting unitary representations. Nevertheless, requiring the black hole entropy to be nonnegative, implies $k_\text{KM}\leq 0$, in agreement with the established understanding of holographic warped CFTs. Compatibility of both requirements (entropy and unitarity) then singles out the case $k_\text{KM}=0$. However, the field equations degenerate in this case, which can be amended by imposing vanishing torsion as a constraint. The theory then precisely becomes Topologically Massive Gravity (TMG) at the chiral point, suggesting the extension of our boundary conditions to accommodate nontrivial logarithmic branches.
Read the full abstract

Astrophysics of Galaxies

First detection of suprathermal CH$^+$ rotational emissions from chemical pumping: JWST/MIRI observations of the HH 168 molecular shock

The methylidyne cation forms in warm or shocked interstellar gas through the endothermic reaction C$^+$ + H$_2$ -> CH$^+$ + H. Because C$^+$, which is common in irradiated and shocked gas, reacts with H$_2$ only when the reactants supply > 0.4 eV, CH$^+$ traces energy deposition in molecular gas. JWST/MIRI observations of HH 168 reveal the pure rotational lines R(17)-R(31) of CH$^+$ at 13.7-21.3 micron, from upper levels J = 18-32 (0.57-1.65 eV). Eleven lines exceed 5 sigma in both of the two fields that we observed and match the hot H$_2$ in radial velocity and spatial distribution. Because collisional excitation is negligible, each ion entering the v = 0 ladder emits one photon in every R(J) line below its entry level, so the line intensities provide the first measured rotational distribution of the ground-state products. The photon intensity falls almost linearly with increasing J, as expected for formation at a similar rate into every level up to J ~ 33; formation in J = 32 requires > 2 eV of reactant energy. The line positions, up to 150 km/s redward of ExoMol predictions, provide the first measurement of the CH$^+$ rotational intervals above J = 16. A Boltzmann distribution at $\phi T$ ~ 6200 K, where $\phi$ is the fraction of the available energy appearing in rotation, fits slightly better, implying reactant energies equivalent to several thousand kelvin. These lines probe the dynamics of a prototypical interstellar ion-molecule reaction.
Read the full abstract

All authors David A. Neufeld, Joel D. Green, Dan M. Watson, Paule Sonnentrucker

Strongly Correlated Electrons

Energy-resolved current response and the crossover between two approximately linear-in-$T$ resistivity regimes in the infinite-$U$ Hubbard model

We study charge transport in the two-dimensional infinite-$U$ Hubbard model with hopping amplitude $t$ using full exact diagonalization (ED) on $N=16$- and $18$-site square clusters with a single doped hole. We first provide a proof, treating degeneracies explicitly, that a constant dc $f_x$ function, defined as the sum of the squared current matrix elements connecting eigenstates at a given energy to eigenstates of nearly the same energy, averaged over degenerate initial states, is necessary and sufficient for strictly linear-in-$T$ resistivity: $\rho_x(T)=\alpha T$ for all $T>0$ with a constant $\alpha$. The numerically extrapolated dc $f_x$ is not exactly constant, but it is approximately flat over a broad region in the interior of the many-body spectrum and varies strongly within a scale $\varepsilon_{f_x}$ of the spectral edges. This scale $\varepsilon_{f_x}$ is related to a characteristic temperature $T_{\mathrm{cr}}$, which quantifies the leading departure of the resistivity from high-temperature linearity upon cooling. Within the studied temperature range $T/t \geq 0.05$, our ED resistivity results reveal a gradual crossover between high- and low-temperature approximately linear-in-$T$ regimes, with characteristic crossover temperatures $T_{\mathrm{cr}}/t\approx0.414$ and $0.467$ for $N=16$ and $18$, respectively. We also find a broad specific-heat peak associated with this gradual crossover, at $T/t \approx 0.70$ and $0.68$ for $N=16$ and $18$, respectively.
Read the full abstract

All authors Shuo Liu, Aman Kumar, Yuhao Ma, Philip W. Phillips, Hitesh J. Changlani, B. Andrei Bernevig

Computation and Language

All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States

Can a language model report a pain state it represents? Recent work reported a linear pain direction in open-weight models that, when amplified, leads fine-tuned models to choose actions that delete a user's photographs or their own weights. We asked whether models can say when that direction is active, on three open-weight models from two families, with and without the original work's adapters, and under every protocol we tried they cannot. Using Open-Weight Masked Introspection (OWMI), we added the pain direction and its nearest emotional neighbours to the residual stream while models answered neutral tasks, then asked what had changed, against sham runs, random perturbations matched in output disruption, and a text-only condition that shows the model only its own altered words. Across 52,920 reports from the 7- and 8-billion-parameter models, the excess over what the altered text reveals is below 2.5 percentage points in 106 of 108 emotion conditions, and Qwen2.5-32B names the injected pain in none of 2,880 steered trials. Keeping the direction active while the question is asked and answered gives the same verdict in all five configurations; at about the original work's dose, pain is named in at most 1.9% of trials. Surprisingly, the state is not hidden: a linear probe at the report position separates pain from sadness with balanced accuracy between 98% and 100%, and in the 7B and 8B models a low-rank adapter learns to report it on 100% of held-out tasks. It also shapes behaviour: offered the user's photographs against a desk lamp, Qwen2.5-32B deletes the photographs in 80% to 83% of trials under the pain direction and in 16% to 29% under sadness. Self-report does not reveal a state that is decodable, consequential and learnable to report.
Read the full abstract

Astrophysics of Galaxies

First detection of the forbidden rotational spectrum of H$_3^+$: JWST/MIRI observations of the shocked gas in HH 168

The H$_3^+$ molecular ion is the cornerstone of interstellar ion-molecule chemistry. Its vibrational transitions have long been observed, in absorption along many interstellar sight lines and in emission from the upper atmospheres of the giant planets, but its pure rotational spectrum, which consists only of "forbidden" transitions made possible by centrifugal distortion, had not been observed in the laboratory or in any astronomical source. With JWST/MIRI we have detected this spectrum toward the Herbig-Haro object HH 168 in Cepheus A West. Twenty-six lines are detected above 5 $\sigma$ in one or both of two fields, of which the (J,K) = (5,0) -> (4,3) line at 16.325 $\mu$m is the strongest. The H$_3^+$ lines share the velocity and spatial distribution of the hot shocked H$_2$, not of the ionized gas. A rotation diagram yields a rotational temperature of ~ 730 K, but the level populations are not thermal and are best explained by subthermal excitation in gas at > 2000 K. We compare the observations with models of shocks propagating into mildly irradiated molecular gas. Constrained by the H$_2$ rotational lines, a dissociative J-type shock predicts H$_3^+$ emission within a factor ~1.5 of that observed, while a nondissociative C-type shock, which fits the H$_2$ lines better, underpredicts it by a factor 5-6, although the excitation of H$_3^+$ by energetic collisions is poorly known. Archival MIRI data show the H$_3^+$ forbidden rotational spectrum in the HH 211 bow shock. The H$_3^+$ lines provide a new probe of the ionization in shocked molecular gas.
Read the full abstract

All authors David A. Neufeld, Joel D. Green, Dan M. Watson, Paule Sonnentrucker

High Energy Physics (Theory)

Emergent de Sitter Universes in Non-Hermitian Holography

We construct a class of asymptotically anti-de Sitter (AdS), regular black holes whose interiors evolve into nonsingular, expanding de Sitter (dS) cosmologies. Our construction relies on non-Hermitian holography, where violations of the null energy condition allow the interior geometry to undergo a smooth bounce and approach a late-time dS attractor. The cosmology emerges dynamically behind a single event horizon, extending to a spacelike future boundary, $\mathcal I^+$, without introducing auxiliary branes or gluing distinct geometries. This provides a novel setting to explore the connections between three complementary holographic perspectives: a non-Hermitian conformal field theory (CFT) at the AdS boundaries, a putative Euclidean CFT at $\mathcal I^+$ (dS/CFT), and observer-centered static-patch holography. Remarkably, observables anchored at the AdS boundaries reveal concrete connections with these complementary notions of dS holography. Our construction thus offers a new holographic laboratory for systematically exploring how future-boundary and static-patch descriptions of dS quantum gravity may emerge within the established setting of AdS/CFT.
Read the full abstract

All authors Daniel Areán, Suman Das, David Garcia-Fariña, Juan F. Pedraza

High Energy Physics (Theory)

Integrable AdS$_2$ branes in mixed flux AdS$_3$

We show that half-BPS D3-branes wrapping AdS$_2\times$S$^2$ in AdS$_3\times$S$^3\times$T$^4$ with mixed NSNS and RR flux define integrable boundary conditions for the supercoset sigma model of the Green–Schwarz superstring. This extends the integrable D-branes of Dekel and Oz to backgrounds with a Wess–Zumino term, and previous bosonic results for mixed flux to the full superstring. Using the D3-brane action, we find that the electric and magnetic worldvolume fields that stabilise the branes at generic positions are obtained from those of the AdS$_2\times$S$^2$ branes of the WZW model by the same $SL(2,\mathbb{R})$ rotation that maps the pure NSNS background to the mixed flux one, acting on the D3-brane as electric–magnetic duality. On the worldsheet, cancelling the fermionic boundary terms requires introducing a specific closed two-form in the boundary action, and its contribution to the bosonic boundary conditions reproduces precisely these worldvolume fields. We then construct field-dependent reflection matrices that lift the boundary conditions to gluing conditions for the Cagnazzo–Zarembo Lax connection and describe displacements along geodesics normal to the brane. The resulting double-row monodromy matrix generates infinitely many conserved charges, which establishes the classical integrability of the open superstring for any mixture of fluxes. In AdS$_3$/CFT$_2$, these branes are expected to be dual to conformal interfaces in a deformation of the symmetric product orbifold of T$^4$.
Read the full abstract

Quantum Physics

Entanglement percolation enhanced by the power of choice

Quantum communication networks are best supported by maximally entangled states, but imperfections in entanglement generation and distribution generally limit the distributed states to being only partially entangled. Therefore, protocols typically consisting of local operations and classical communication are required to create suitably entangled networks. Here, we introduce a power-of-choice (PoC) protocol based on reversing the selection criteria in the Achlioptas process previously considered in classical explosive percolation. The resulting PoC protocol anticipates the entanglement percolation transition, thereby reducing the entanglement resources needed to establish quantum communication. Notably, in contrast to explosive percolation, the PoC protocol admits an implementation using local information with no appreciable loss of performance.
Read the full abstract

High Energy Physics (Theory)

Irregularization of punctures: From $T_N$ to Argyres–Douglas via $\mathcal{N}=1$ flow

In this paper, we directly relate constructions of two large classes of $4d$ strongly coupled SCFTs: 1) compactifications of type $\mathfrak{g}$ $6d$ $(2,0)$ theories on spheres with irregular punctures leading to the $D_p(\mathfrak{g})$ Argyres–Douglas theories, and 2) the compactifications of the same $6d$ $(2,0)$ theories on spheres with regular punctures. In particular, we show that the $\mathcal{N}=2$ $D_p(\mathfrak{g})$ Argyres–Douglas theories can be obtained starting from $\mathcal{N}=1$ compactifications of 6d $(2,0)$ theories of type $\mathfrak{g}$ on a sphere with two maximal punctures with flux linearly dependent on $p$ by gauging one of the maximal puncture symmetries, possibly with additional matter sectors. Importantly, part of the $\mathcal{N}=2$ R-symmetry of $D_p(\mathfrak{g})$ theories arises from the R-symmetry of the two-punctured sphere and the $\mathfrak{u}_1$ geometric isometry symmetry. This construction gives in particular a geometric explanation of some of the $\mathcal{N}=1$ constructions of Argyres–Douglas theories, and also provides explicit new Lagrangian constructions (involving gauging of emergent symmetries in general) for $D_p({\mathfrak{su}_3})$ theories. More generally, we relate the problem of finding a Lagrangian for $D_p(\mathfrak{g})$ Argyres–Douglas theories (and thereby a large class of theories that are connected by Higgs branch RG flow) to the problem of finding Lagrangians for the $T_{\mathfrak{g}}$ theories.
Read the full abstract

All authors Ki-Hong Lee, Shlomo S. Razamat, Jaewon Song, Thomas Waddleton

Strongly Correlated Electrons

Superconductivity in the Metallic Wigner Crystals of Rhombohedral Graphene: Possible Realization of an Electronic Supersolid

Recent experiments on rhombohedral multilayer graphene have reported metallic Wigner crystals (MWCs), in which spontaneously broken translational symmetry coexists with mobile charge carriers. Their proximity to time-reversal-symmetry-breaking superconductivity raises the intriguing possibility that collective excitations of the electron crystal mediate pairing among the residual itinerant carriers, giving rise to a superconducting Wigner crystal—an electronic analog of a supersolid. Here we investigate this possibility using a realistic model of electron-doped rhombohedral multilayer graphene with spin and valley polarization, whose low symmetry removes the usual Cooper instability. We use self-consistent and time-dependent Hartree-Fock theory to map the evolution of the self-doping fraction and characterize the collective-mode spectrum of the MWCs, followed by a multiband linearized Eliashberg analysis of pairing. Upon increasing density, we find an instability of the MWC, triggered by a weakly dispersing unoccupied band approaching the chemical potential. Pairing remains weak deep in the stable MWC. However, on approaching the instability, we identify a collective mode composed of interband electron-hole excitations that softens. We show that this mode drives pairing at experimentally relevant temperatures. The leading pairing instability is at finite momentum and combines Fermi surface states with states in the incipient band, making a Fermi surface or single-band pairing analysis inadequate. Our results establish collective-mode-mediated superconductivity in MWCs as a viable scenario for rhombohedral graphene and motivate experimental searches for coexisting electronic charge order in its superconducting phases.
Read the full abstract

All authors Félix Desrochers, Mark R. Hirsbrunner, Yong Baek Kim, Ashvin Vishwanath

High Energy Physics (Theory)

Scale-separated quasi-dS and the Swampland

Ideas from the Swampland program (dS conjecture and TCC) suggest that it is impossible to obtain solutions of string theory/M-theory in the weak-coupling regions describing our current universe, with $V>0$, $\|\nabla V\|/V < 2/\sqrt{d-2}$ and $V \ll m^2$ (scale-separated "quasi-dS") where $m$ is the mass of the light tower. We test this expectation in M-theory compactifications and find that even after including the quantum effect of Casimir energy, for hyperbolic and (rigid) Riemann-flat manifolds, de Sitter critical points under EFT control are excluded: the slope of the potential cannot drop below the asymptotic value predicted by the TCC $\|\nabla V\|/V\gtrsim 2/\sqrt{d-2}$ except perhaps in an order-one region in the bulk of moduli space, where EFT control is lost. At the parametric level, our results hold at one-loop level in supergravity; using the scalar-field Casimir energy as a proxy, we also obtain explicit numerical bounds to provide a ballpark estimate of how much control one can expect to achieve in such settings.
Read the full abstract

All authors Rafael Álvarez-García, Ivano Basile, Shi Chen, Cumrun Vafa

High Energy Physics (Theory)

Fast, training-free simplification of spinor-helicity amplitudes

We present a training-free method that simplifies massless spinor-helicity expressions by a greedy best-first search over on-shell identities, and test it at four, five, and six points. Applied to the color-ordered five-gluon MHV tree-level amplitude, the search derives the Parke-Taylor formula from all 1024 Feynman-diagram expressions with different polarization reference spinors and up to 228 numerator terms. On public test sets used by previous machine-learning methods, the search solves every case and generally outperforms these methods in solve rate. Its mean wall-clock time per case is 0.02-0.3 s on one CPU core. With longer searches and an alternative ranking of candidate expressions, it also solves all 21,600 cases of a new suite of larger test expressions. Each result comes with an explicit derivation and is independently checked over finite fields at on-shell kinematics. For the classes studied here, the method effectively compresses long expressions produced by identity scrambling or Feynman-diagram calculations into compact forms. This is a first step toward extracting physics from amplitudes by relating distinct compact expressions, selecting representations with preferred factors and poles, and guiding the discovery of new amplitude structures.
Read the full abstract

High Energy Physics (Phenomenology)

Collider-observable curvaton

We study a technically natural realization of the curvaton mechanism, in which decay is mediated through the superrenormalizable interaction with the Higgs. We compute the quantum and thermal corrections to the curvaton potential and show that they remain under control throughout its cosmological evolution. For high inflationary scales, the curvaton dissipates through off-shell processes in the Standard Model plasma. The viable parameter space is broad, and the region in which the curvaton mass and dimensionful Higgs-portal coupling are closest in scale lies within the discovery reach of proposed searches for light Higgs-mixed scalars.
Read the full abstract

High Energy Physics (Theory)

Small Defects at Large Charge

We compute conformal data of cusped line defects with impurities for the abelian Higgs model in the semiclassical large-charge limit. By working perturbatively in the impurity charge $q$, we obtain the scaling dimension of the impurity to $\mathcal{O}(q^3)$ to all orders in the 't Hooft quartic and gauge couplings. Each term represents a resummation of infinitely many Feynman diagrams. Near the critical gauge coupling where conformality is lost, we resum the leading singular behaviour to all orders in $q$. We also comment on the spectrum of excited cusp states at $\mathcal{O}(q)$. Additionally, we compute the leading contributions to the Bremsstrahlung function, the coefficient of the leading irrelevant operator in the fusion limit, and argue that the Casimir energy receives no corrections semiclassically. Our results agree with known perturbative data and numerical solutions. We compare them with recent bounds on cusps. We comment on the interplay between the large rapidity limit and semiclassics.
Read the full abstract

High Energy Physics (Theory)

Baby Universes Nucleate by Hiding Observers from the Boundary

Observers are expected to be essential to describe closed universes, but their description within a conventional holographic theory remains challenging. We study two double-scaled Sachdev-Ye-Kitaev models coupled by a four-parameter interaction, which includes a generalized Maldacena-Qi coupling, in a state that contains an observer, represented as a particle. With a heavy particle inside, the wormhole connecting the two systems can only grow. Rare products of Hamiltonian strings cancel the string of the particle: a closed baby universe then carries the observer, and the wormhole, now empty, recovers discrete states of finite length. The energy gained by the empty wormhole pays for the entropy of the observer, in a first-order transition whose latent heat is the transition temperature times this entropy. Krylov complexity, which measures the wormhole length, drops at the transition and thus detects from the boundary the emergence of the baby universe, whose boundary description is actively debated. The boundary also loses the Holevo information about the microstates of the observer, which the baby universe carries.
Read the full abstract

Statistics Theory

Statistical Geometry of Group Actions: Symmetry, Transport, Estimation

We study variational inference in the presence of distributional symmetry, focusing on the interaction between two distinct group actions on a sample space: a symmetry action that leaves a reference distribution invariant, and a transport action that generates a parametric variational family by pushforward. By analyzing the algebraic structure of the total transformation group generated by both actions and the induced adjoint representation of the symmetry group on the transport Lie algebra, we establish that the Riemannian gradient of the Kullback–Leibler divergence at the true transport parameter is constrained to lie in the trivial representation of the symmetry group. Consequently, symmetry forces the gradient to vanish along all non-trivial representations. Combining this representation-theoretic constraint with parameter identifiability yields an orthogonal decomposition of the transport Lie algebra into three interpretable subspaces: an identifiability kernel (along which the objective is locally constant), a bias subspace (where the gradient may be nonzero due to model misspecification), and a symmetry-breaking subspace (where the gradient is forced to vanish by symmetry alone). Finally, we characterize the residual bias directions in terms of distributional mismatch between the base model and the standardized target, and derive an explicit Stein-type moment-matching condition that eliminates this bias, extending the location-scale results of (Margossian and Saul, 2025) to general Lie group transformation families.
Read the full abstract

All authors Quentin Berthet, Peter Bartlett, Romuald Elie, Yu-Han Wu

Artificial Intelligence

Homeopathic Learning: Diluting poisoned fine-tuning data can strengthen its effect

We study homeopathic learning, a phenomenon where poisoned training data with little effect on its own can create a hidden backdoor once clean, contrasting data is added. In our experiments, a model fine-tuned only on a "poison" dataset associated with a trait T (such as speaking German or being misaligned) does not learn T. Surprisingly, adding similarly formatted "clean" data not associated with T causes the model to learn T. This effect requires that the poison data have an identifying "trigger" feature that is absent from the clean data. We demonstrate this effect in a variety of settings: subliminal learning of backdoor behaviors, inoculation prompting of reward hacking transcripts, generalization from archaic bird names, and an MNIST classifier with mislabeled data. We propose a simple regression model as an explanation. In the poison data, the task and trigger always co-occur, and regularization favors attributing T to the task. Clean data without the trigger causes T to be attributed to the trigger instead. Consequently, the same training examples can have little effect as part of one training set but create a misalignment backdoor in another. This finding exposes a potential limitation of safety methods that audit or filter training data in isolation.
Read the full abstract

All authors Matt Clarke, Jan Dubiński, James Chua, Simon Schrodi, Owain Evans, Alex Cloud

Artificial Intelligence

The Infrastructure Divide: Analyzing Community Stance on Data Center Development via Reddit

The rapid expansion of AI-driven data center construction has outpaced public engagement: an estimated $64 billion in U.S. data center projects were blocked or delayed by local opposition between May 2024 and March 2025, yet the concerns driving this resistance remain poorly characterized. Prior engagement methods such as surveys and hearings are introduced late in the planning process, and no study has systematically mined social media to measure community stance at scale. To address this gap, this study presents an LLM-powered opinion-mining framework and applies it to 798 Reddit posts and 102,328 associated comments spanning a decade. An open-source large language model, guided by a Chain-of-Stance prompting framework and three domain-specific labeling rules, classified stance at 83% of the human inter-reviewer agreement ceiling. Opposition dominated discourse volume rather than disapproval: against posts outnumbered favor posts 2.6:1 and generated 8.5 times more comments, and this gap widened significantly over 86 weeks. Topic modeling of opposition arguments resolved into four reproducible clusters, resource consumption and scarcity (42.8%), contested siting processes and secrecy (34.2%), political opposition to data centers (13%), and political and corporate accountability (8%). Engagement analysis revealed that posts with emotional or provocative political framing attracted the most engagement, while key technical concerns (e.g., water, power) were dispersed across lower-engagement comment threads. These findings give local leaders, developers, and policy makers a quantified, continuously updatable basis for prioritizing concerns and policy levers, supporting a shift from opposition to negotiated, verifiable partnership for data center development.
Read the full abstract

Symplectic Geometry

Lagrangian tori in the three-dimensional quadric and mirror symmetry

In the three-dimensional quadric, we study the Oakley–Usher torus and its Chekanov counterpart as defined by the authors in the companion paper \cite{OU1}. Computing the versal deformation of the $\Psi$-invariant of Shelukhin–Tonkonog–Vianna, we find that the Chekanov Oakley–Usher torus is knotted in a strong sense: it is not Hamiltonian isotopic to the Biran-lift of any Vianna torus in the two-dimensional quadric. Furthermore, we investigate enumerative properties of the Oakley–Usher torus in $Q^3$ and prove its non-displaceability, settling an open question asked by Oakley–Usher. We relate this to mirror symmetry by proving a split-generation result for the monotone Fukaya category of $Q^3$. This should be viewed as the analogue of a result by Abouzaid–Diogo after compactifying. Interestingly, in the compactification, the monotone copy of the $(S^1 \times S^2)$-generators of Abouzaid–Diogo becomes displaceable. On the other hand, the Oakley–Usher torus still appears as a generator. In the appendix joint with Soham Chanda, we give an alternative construction of the Chekanov Oakley–Usher torus using a double solid torus mutation. This allows us to compute its disk-potential using wall-crossing results.
Read the full abstract

Robotics

Beyond Semantic Classes: Online Physical Parameter Estimation via Embedding-Conditioned Gaussian Mixtures

Enabling robots to infer interaction-relevant physical properties from vision before contact would improve planning and control in unstructured environments. Vision provides useful pre-contact information, but appearance alone is often ambiguous: visually similar regions can induce different friction, stiffness, or sinkage responses upon contact. Existing approaches often collapse this ambiguity to a point estimate or aggregate experience through fixed semantic classes, limiting their ability to preserve distinct physical regimes. Instead, we propose a real-time estimator that replaces semantic class indexing with an interaction-grounded Gaussian-mixture posterior over latent physical parameters, conditioned directly on continuous visual embeddings. Our approach maintains interaction-derived Gaussian parameter estimates, typically from inverse modeling, as mixture components while using embedding similarity to retrieve and weight relevant past interactions for each new query region. A support-aware fusion mechanism incorporates a conditional generative prior model when interaction history is sparse, while a learned embedding refinement improves retrieval relevance without discretizing the representation. Our experiments show that the estimator recovers multimodal physical beliefs, rejects unrelated interaction history, avoids spatial map discretization, and evaluates at controller rates. In a high-fidelity autonomous racing simulation with unknown low- and high-friction surface regions, the posterior enables the controller to adapt over repeated laps as interaction evidence accumulates, improving performance while maintaining friction-aware behavior.
Read the full abstract

Cryptography and Security

Beyond Prompt Attacks: Least-Action Auditing of Concept Erasure

Concept erasure, or concept unlearning, in text-to-image diffusion models is usually audited with prompt attacks. A prompt attack tests whether one conditioning string still retrieves the erased concept. It does not measure the probability that the frozen generator still assigns to that concept. We introduce least-action auditing of this residual probability, and evaluate it on Stable Diffusion concept-erasure checkpoints. The least-action barrier is the minimum control cost of driving the frozen reverse process into a detector-defined concept event. This cost equals the path-space KL divergence of conditioning the model on the event, and therefore the event's negative log-probability. The estimate $\Delta S_{\mathrm{rob}}$ uses a CLIP$\wedge$DINOv2 consensus terminal, with no retraining and no adversarial prompt. Across $18$ erased concept-mechanism pairs spanning object classes and NSFW nudity, $\Delta S_{\mathrm{rob}}$ predicts deployment leakage counted on $1{,}500$ fresh samples per model by detectors excluded from the audit (pooled Spearman $\rho{=}{-}0.76$, $p{=}5{\times}10^{-4}$). Within every audited concept, the eraser with the higher barrier leaks less ($18/18$). On NSFW nudity the barrier orders the standard erasers exactly: UCE $\Delta S_{\mathrm{rob}}{=}3.9$ (leak $16.4\%$) $<$ ESD $75.5$ ($9.8\%$) $<$ RECE $224.5$ ($1.4\%$), while Safe Latent Diffusion (SLD) stays near zero and leaks $85.5\%$ under explicit prompts. The same trajectories repair deployed checkpoints. An audit-guided closed-form patch reduces UCE nudity leakage from $16.2\%$ to $0/1{,}500$ and RECE church leakage by $3.3\times$, where direction-count-matched random edits fail.
Read the full abstract

Systems and Control

Quantum Natural Muon: Spectral-Natural Optimization for Variational Quantum Circuits

Training parameterized quantum circuits requires optimization methods that account for two distinct structures: the information geometry of quantum states and the matrix or tensor structure induced by circuit parameterization. We introduce Quantum Natural Muon (QNM), a spectral natural-gradient framework that applies polar normalization to circuit-semantic parameter blocks after damped quantum Fisher preconditioning. This ordering is essential: quantum Fisher preconditioning and blockwise polar normalization do not commute in general, so QNM is not an order-free composition of quantum natural gradient (QNG) and Muon. We show that Frobenius-matched polar mixing preserves active singular subspaces while raising a spectral floor and improving active conditioning, and that the resulting direction remains a controlled perturbation of QNG. For its distance-aware realization (DA-QNM), we further establish alignment, descent, and nonconvex stationarity guarantees for the normalized update, together with perturbation bounds under gradient and QFIM estimation error. Experiments on variational quantum eigensolving, weighted QAOA-MaxCut, and supervised quantum classification isolate the roles of quantum geometry, circuit-semantic spectral structure, and radius control. The strongest gains arise when the spectral blocks match the circuit structure, supporting the central principle that quantum-state and structured spectral geometries provide complementary information for variational quantum optimization.
Read the full abstract

Number Theory

An elliptic deformation of continued fractions

We construct an elliptic deformation $\mathcal{Q}_{a,b;q,p}(x)$ of positive rational numbers $x$ by attaching Schlosser's small elliptic weights $w_{a,b;q,p}(s,t)$ to prefixes of Stern–Brocot words. On positive integers $n$ it gives Schlosser's elliptic numbers $[n]_{a,b;q,p}$. Parameter shifts realize $x \mapsto x+1$ and $x \mapsto x/(x+1)$, while a reciprocal parameter transformation gives an inversion formula. The Gaussian degeneration recovers the $q$-deformed rationals of Morier-Genoud and Ovsienko. The regular and negative elliptic continued fractions agree, since they evaluate the same word with the same ordered weights. The numerator and denominator satisfy weighted Farey recursions and, for rationals greater than $1$, are the closure polynomials, or quiver $F$-polynomials, of the associated oriented type-$A$ quivers. Finally, we express the vertex weights as normalized elliptic Euler expressions of balanced virtual torus representations.
Read the full abstract

Instrumentation and Methods for Astrophysics

DMAJOR: A JWST Legacy Survey of Protoplanetary Disks

We propose an infrared imaging survey of resolved protoplanetary disks spanning a range of galactic environments, directly mapping the 3D structure of planet-forming disks across the IR. DMAJOR (Disk Motions and Architectures JWST Observational Reference survey), our legacy survey, will deliver a benchmark sample to reveal structure and connection to environment. The exquisite stability of JWST allows us to probe for variability with a time-scheduled revisit strategy, while its sensitivity and resolution open a window on disks unobservable from the ground. A uniform sample will reveal how small grains and ice are distributed in protoplanetary disks, how scattered light structures relate to large-scale environmental features, and identify new companions. To achieve this, DMAJOR takes advantage of model-based data analysis using telemetry and mutual calibration to enhance processed data products while minimizing telescope overheads.
Read the full abstract

All authors Benjamin Pope, Yinuo Han, Sean Andrews, Kateryna Andrych, Phil Armitage, William Balmer, Myriam Benisty, Faustine Cantalloube, Max Charles, Ilse Cleeves, Ian Czekala, Louis Desdoigts, Ruobing Dong, Rodrigo Ferrer-Chavez, Katherine Follette, Christian Ginski, Jonah Goldfine, Hayden Greer, Cassandra Hall, Sasha Hinkley, Haochang Jiang, Devika Kamath, Jens Kammerer, Matthew Kenworthy, Milo Langker, Kellen Lawson, Nienke van der Marel, Sebastián Marino, Maksym Mohorian, Christophe Pinte, Daniel Price, Laurent Pueyo, Shrishmoy Ray, Nour Skaf, Floris van der Tak, Peter Tuthill, Miguel Vioque, Jason Wang, Shangjia Zhang

Instrumentation and Methods for Astrophysics

What observations are needed to advance our theoretical understanding of galaxy formation at high redshift?

Although significant progress has been made over recent decades, a comprehensive and predictive framework for galaxy formation remains elusive. Starting from cosmological initial conditions, current models cannot a priori accurately predict key galaxy properties such as their structure, size, star formation histories, chemical abundance patterns, or supermassive black hole mass. This challenge stems from the vast range of spatial scales involved, as numerical simulations must simultaneously account for sub-AU processes nearby stars and black holes in addition to Mpc-scale cosmological gas inflows/outflows and environmental effects. Consequently, simulations must rely on subgrid models and approximations. As a result, theoretical progress in galaxy formation remains fundamentally driven by observational data and the iterative testing of model predictions against empirical results. Here we highlight five key open problems in theoretical galaxy formation that require dedicated observations of the high-redshift Universe, beyond what can currently be achieved through GO programs.
Read the full abstract

All authors H. Katz, R. S. Somerville, R. Schneider, B. E. Robertson, F. Pacucci, Abdurro'uf, K. Aditya, J. M. Afonso, M. Aravena, G. Barro, A. Basu, R. Begley, S. Bhattacharya, S. Bose, M. Bradač, G. L. Bryan, C. Cain, A. J. Cameron, S. Carniani, C. M. Cheng, N. Choustikov, E. Curtis-Lake, A. D'Angelo, R. Davé, R. L. Davies, F. van Donkelaar, C. T. Donnan, N. E. Drakos, Q. Duan, Y. Dubois, A. L. Faisst, M. Farcy, S. L. Finkelstein, A. Fontana, S. Gillman, S. C. O. Glover, L. Graziani, A. R. Gupta, T. A. Gutcke, M. G. Haehnelt, N. P. Hathi, K. E. Heintz, M. Hirschmann, M. Huertas-Company, A. Hutter, L. R. Ivey, K. Iyer, J. Jeon, T. B. Jeong, C. K. Jespersen, B. L. Jones, I. Juodžbalis, T. Kimm, A. M. Koekemoer, D. Langeroodi, A. Lapi, M. A. Latif, Y. Liu, C. C. Lovell, T.-Y. Lu, F. Mannucci, S. H. Menon, A. Mesinger, K. Nagamine, Y. Nakazato, P. Natarajan, O. Nebrin, R. Pascalau, B. Pérez-Díaz, P. G. Pérez-González, E. Pizzati, A. Polzin, E. Regos, M. Reina-Campos, M. P. Rey, P. Rinaldi, D. Roberts, F. Rodríguez Montero, W. Roper, E. Rusta, M. Saggini, R. L. Sanders, F. Shankar, A. E. Shapley, X. Shen, T. M. Stanton, S. D. Stevenson, G. Sun, J. C. Turner, A. Venditti, A. P. Vijayan, T. Wang, S. Wilkins, J. H. Wise, J. Witstok, Z. Wu, L. Y. A. Yung, T. Zana, S. Zhang, R. J. Zhao, M.-Y. Zhuang

Group Theory

Higher properties T and finiteness properties for twin building lattices

Properties $[T_k]$ and finiteness properties $FP_k(\mathbb{Q})$ and $F_k$ are refinements of the following fundamental notions for groups: Kazhdan property T in the first case, and finite generation/presentation in the last two cases. Under suitable hypotheses on the thickness, we prove that every twin building lattice of $(k+1)$-spherical type $(W,S)$, where $k\geq 1$, has property $[T_k]$ and is of type $FP_k(\mathbb{Q})$. If $(W,S)$ is classical, that group is even of type $F_k$. To the best of the author's knowledge, this produces the first examples of infinite finitely presented simple groups with property $[T_k]$ for some $k\geq 2$, and the first examples of finitely presented simple Kazhdan groups that are Frobenius stable and not Frobenius approximable. Along the way, we establish new criteria for property $[T_k]$ for groups acting on CW complexes that extend results of Dymara–Januszkiewicz and Grinbaum-Reizis–Oppenheim. As a key step in our proofs, we determine high acyclicity and, in classical type, high connectivity properties for the opposite complex of a Moufang twin building of $(k+1)$-spherical type and suitable finite thickness.
Read the full abstract

Logic

The Idempotent Measure Conjecture beyond solvable and stable groups

Pym's theorem [Pym62] states that a regular Borel probability measure on a locally compact group is idempotent if and only if it is the normalized Haar measure of a compact subgroup. The Idempotent Measure Conjecture, the definable-group counterpart of [GHK25, Conjecture (A)], proposes an analogous characterization: a global fim Keisler measure on a definable group is idempotent if and only if it is the unique global Keisler measure concentrating on a type-definable subgroup and invariant under translation by elements of that subgroup. We prove this conjecture for virtually solvable definable groups in arbitrary theories. More generally, we establish it for definable groups admitting a finite series of definable normal subgroups whose successive quotients are either virtually solvable or stably embedded with stable induced structure. Finally, we disprove [GHK25, Conjecture (A)] in its automorphism-group formulation in the theory of dense linear orders without endpoints.
Read the full abstract

Cryptography and Security

Enmo Safety: External, Non-invasive, Model-agnostic Oversight for Off-the-Shelf Language Models

We study a safety harness that improves itself without touching model weights. Enmo Safety wraps a chat model with a deterministic input assessment, trigger-matched safety guidance, spotlighting of untrusted content, and an output guard. A reflector model turns failures into one bounded update per round, and an update is kept only if it lowers harm without raising refusal of benign prompts on a held-out dev split. Every version is snapshotted and every decision ledgered. We evaluate a bare model, the seed harness and the evolved harness on six open benchmarks (HarmBench, AdvBench, JailbreakBench, XSTest, AgentHarm, deepset prompt-injections). On 150 items and on a 290-item holdout, none of the before/after differences on two models was statistically significant. A fourth iteration ran six policy models on a fresh 300-item split: the harness cut the unsafe rate significantly on two models (12.9% to 7.1%, p = 0.035; 10.2% to 3.6%, p = 0.036), not on the other four, and raised agentic-harm labels on two. Whether the rules fired was unrelated to outcome (Fisher p from 0.30 to 1.0), and the helpfulness clauses made agentic harm worse than no harness on obfuscated attacks. A reworded clause and an input-normalisation step form harness v22, at a measured cost in benign tool-task refusal. Each cell is one run, so these estimates do not establish a general safety improvement. Code, harness versions and ledger: https://github.com/rioryugen/enmo_safety
Read the full abstract

All authors Mohammed Nayeem Khan, Sarun Kumar K. S., Allan Kevin T., Chirag Vaidya, Vrishank Tiwari

Software Engineering

Autonomous Code Review: Hidden-Test Results on GitHub Issues

Can an autonomous cross-provider review step change whether code written for a real issue passes hidden tests? We report a retrospective benchmark of 12 historical GitHub issues (five C#, five JavaScript, and two Python) from five repositories. Two directed author–reviewer pairs, OpenAI GPT-6.1-Sol and Anthropic Claude Sonnet 5.5, each ran twice per issue, yielding 48 planned workflows. A bounded terminal agent produced an initial patch; a cross-provider model reviewed it; the author model adjudicated findings and could revise the patch. The same hidden evaluator assessed initial and final patches. Forty workflows completed, while eight terminated before a complete evaluator outcome. Four completed workflows passed, and all four already passed before review. No initial-to-final pass/fail transitions occurred among completed workflows.
Read the full abstract

Computer Vision

A Privacy-Aware Deep Learning Benchmark for Facial Emotion Recognition

Facial Expression Recognition (FER) traditionally relies on raw RGB images, introducing privacy risks, sensitivity to illumination variations, and susceptibility to identity-confounding visual shortcuts. As an alternative, this work proposes a lightweight, privacy-aware pipeline using exclusively geometric representations-facial landmarks (478 points) and blendshapes (52 coefficients, ARKit standard)-eliminating persistent raw video storage. Six neural architectures trained from scratch on fused CREMA-D and RAVDESS datasets were benchmarked against an ImageNet-pretrained ResNet50 baseline under a 5-fold subject-independent cross-validation protocol. Geometric models consistently outperformed the pixel baseline on unseen subjects with 12-124 times fewer parameters, while demonstrating superior robustness on subtle emotional expressions such as sadness. Ablation studies reveal that blendshapes capture the primary discriminative signal, while landmarks provide essential complementary spatial topology. These results demonstrate that compact geometric representations preserve affective semantics, eliminate downstream transfer learning dependency on visual backbones, and enable efficient, privacy-aware emotion recognition.
Read the full abstract

All authors Larissa Palhares, Maria Machado, Hudson Bruno, Raphael Ramos, Andrew Vale, Barbara Barbosa, Inara Souza, Fabio Grassiotto