Computer Vision
MIRAGE: Metric Inference and Reliability-Aware Gaussian Estimation from Feed-Forward Vision
Xinzhe Wang, Changjian Jiang, Kaiwen Song et al. (+8 more)
Feed-forward 3D priors enable high-quality reconstruction from image streams but often lack reliable metric scale and suffer from long-term trajectory drift. In contrast,
LiDAR–inertial estimation provides metrically accurate motion but lacks the image-space consistency needed for high-fidelity reconstruction. We propose $MIRAGE$, which anchors feed-forward visual geometry to
LiDAR–IMU metric motion and regulates online pose corrections according to correspondence support and cross-modal agreement, jointly enabling accurate metric pose estimation and high-fidelity Gaussian reconstruction. Experiments on Oxford Spires, M2DGR and INS show that MIRAGE achieves accurate metric trajectory estimation, outperforming COLMAP, a widely used but computationally expensive reference for
vision-only
benchmarks. Motivated by the scarcity of long-horizon
monocular benchmarks for large-scale indoor reconstruction, we further build a
multimodal data-collection platform and use MIRAGE to establish accurate 6-DoF reference trajectories for a new
vision-only
benchmark. We
benchmark recent
monocular large-scale reconstruction methods on this
dataset, providing a systematic evaluation of their performance in long-horizon indoor scenes.
Read the full abstract ▾
All authors Xinzhe Wang, Changjian Jiang, Kaiwen Song, Xudong Li, Kerui Ren, Ran Yi, Lizhuang Ma, Chunhua Shen, Linning Xu, Tao Lu, Mulin Yu
Robotics
CASTER: Formalizing Zero-Shot Manipulation from Synthetic Videos as Constrained Optimization
Yinlong Dai, Homanga Bharadhwaj and Dylan P. Losey
Prior research on learning from humans often requires real human demonstrations. But advances in
video generation models offer a new paradigm: we can now
prompt these generators to make synthetic
videos of the task. On the plus side, these synthetic
videos take place in the
robot's current environment, and do not require actual human teaching. However, while the generated
videos appear to show plausible motions, they lack both physical consistency and
robot grounding, and thus the
robot cannot directly roll out the trajectory shown by the
video. In this paper we seek to formalize the process of going from synthetic
video generation to real-world
robot trajectories for zero-shot manipulation tasks. We recognize that this problem is an instance of constrained optimization. The object trajectories extracted from the synthetic
video provide an initial guess at the correct motion, and the
robot must refine these trajectories (for instance, within a digital twin) while enforcing physical feasibility and semantic consistency. Our optimization-based formalism provides insights into the
video generation properties necessary for success. Specifically, we show that the
video generation must probabilistically seed the search within a semantic basin of attraction where the trajectory can be locally modified without losing its task
alignment. Our analysis further reveals that state-of-the-art approaches can be viewed as instances of our unified formalism with key simplifications. In our experiments we remove these simplifications to more closely match the underlying optimization, leading to more robust zero-shot trajectories from synthetic
videos and improving average task success by 34.1% compared to the best-performing baseline. See
videos at our project website: https://collab.me.vt.edu/CASTER
Read the full abstract ▾
Machine Learning
Sample-Adaptive Zeroth-Order Post-Training
Haiyuan Wan, Yanyi Li, Weicheng Lin et al. (+1 more)
Zeroth-order (ZO) optimization offers a
memory-efficient alternative to backpropagation for fine-tuning large
language models. However, existing ZO methods typically sample training examples uniformly, overlooking their evolving importance and leaving avoidable estimation variance. To address this limitation, we propose Sample-Adaptive Zeroth-Order Optimization (SAZO), which learns sample importance online and adaptively allocates queries to accelerate convergence. SAZO extracts gradient-norm information from existing directional differences to estimate sample importance without access to true gradients. Furthermore, to track evolving importance from sparse minibatch feedback, SAZO maintains per-example exponential moving averages that aggregate observations across updates. The method requires no additional model evaluations and only $O(N)$ additional scalar state compared with MeZO. We prove that SAZO asymptotically matches the leading convergence term of the minimum-variance sampling oracle, and achieves a strictly faster convergence rate than standard methods. Extensive experiments across four model families demonstrate improved average downstream performance and up to $6\times$ faster convergence to the same training loss in terms of query count.
Read the full abstract ▾
All authors Haiyuan Wan, Yanyi Li, Weicheng Lin, Cong Fang
Classical Analysis and ODEs
A higher-dimensional Schrödinger maximal estimate along families of bi-Lipschitz curves
Javier Minguillón
Let $\Theta\subset\mathbb{R}^n$, $n\geq 1$ be compact with upper Minkowski dimension $\beta$, and let $\gamma(x,t,\theta)$ be a family of curves which is bi-Lipschitz in the first variable and Lipschitz in the last two. We prove the maximal estimate $$\left\|\sup_{(\tau,\theta)\in[0,1)\times\Theta}\left|e^{i\tau\Delta}f(\gamma(\,\cdot\,,\tau,\theta))\right|\right\|_{L^2(B^n_1)}\lesssim_s\|f\|_{H^s(\mathbb{R}^n)},$$ for every $s>\tfrac{n(\beta+1)}{2(n+1)}$, in the range $0\le\beta\le n-1$. In the range $n-1<\beta\leq n$ we prove the threshold $s>\min\big\lbrace\frac\beta 2 + \frac 1{2(n+1)},\frac n2\big\rbrace$. The proof relies on a fractal $L^2$ estimate of Du–Zhang. We also use the temporal localization of Lee, the locally constant property, and a counting argument allowed by the Lipschitz properties of the curve.
Read the full abstract ▾
High Energy Physics (Phenomenology)
Seeing in the Dark: Gravitational Waves from Resonant Production of Dark Photon Dark Matter
Yanou Cui, Pankaj Saha and Evangelos I. Sfakianakis
Ultralight dark
photons (DPs) are well-motivated dark matter candidates. A promising yet underexplored mechanism for their relic abundance is parametric resonance in the Abelian-
Higgs model that generates the DP mass. We advance previous analytical studies through dedicated 3D
lattice simulations that capture important nonlinear effects, including backreaction and relic partitioning between the dark
Higgs and DPs. We further predict the associated
stochastic gravitational-wave (GW) signal, which can be within reach of a broad range of detectors, opening a new avenue to probe ultralight DP dark matter.
Read the full abstract ▾
Superconductivity
Enhanced Coulomb pseudopotentials in high-pressure hydrides from first principles
Kieran Bozier and Chris J. Pickard
In the Eliashberg theory of superconductivity, Coulomb interactions oppose the
phonon-mediated attraction and suppress the superconducting critical temperature $T_{\mathrm{c}}$. However, most first-principles predictions continue to treat this repulsion with an empirical Coulomb pseudopotential $\mu^*_{\text{ME}}$ in the range 0.1-0.15. Here we present first-principles calculations of the Coulomb parameter $\mu$ and pseudopotential $\mu^*_{\text{ME}}$ in superconducting hydrides across a range of pressures. The screened Coulomb interaction is evaluated at the random phase approximation level with the Mantel code. We demonstrate that several hydride systems, including H$_3$S, YSbH$_6$ and Mg$_2$IrH$_6$, have strong Coulomb repulsions driven by a large density of states at the
Fermi energy, $N_{\mathrm{F}}$. For Mg$_2$IrH$_6$ we obtain $\mu=0.62$ and $\mu^*_{\text{ME}}\approx0.39$, leading to a 40 K suppression in $T_{\mathrm{c}}$ relative to a standard $\mu^*_{\text{ME}}=0.16$. Our obtained $\mu^*_{\text{ME}}$ is far above the conventional range and, accounting for the energy dependence of the density of states, resolves the factor-of-two discrepancy between $T_{\mathrm{c}}$ estimates of previous studies. Across all systems and pressures studied, we find that $\mu$ closely tracks $N_{\mathrm{F}}$, suggesting that commonly used $\mu^*_{\text{ME}}$ values lead to overestimated $T_{\mathrm{c}}$ in materials where the density of states at the
Fermi energy is large.
Read the full abstract ▾
Earth and Planetary Astrophysics
The accretionally heated atmospheres of protoplanets and brown dwarfs
Nick Choksi, Peter Gao, Jerry Xuan et al. (+2 more)
The ranks of accreting
planets and brown
dwarfs are growing. To help interpret their spectra, we study how
accretion affects a
planet's atmosphere. The atmosphere is irradiated from above by whatever nonthermal radiation from the
accretion shock does not directly escape into space. This accretional heating drives a temperature inversion. Even relatively slow
accretion, corresponding to mass doubling times longer than system ages, can raise temperatures by many hundreds of Kelvin at ambient pressures of order millibars. The heated atmosphere radiates an optical/infrared excess. For environments that are not too dusty, the excess concentrates in molecular lines that trace the heated layer. Thus an accretionally heated atmosphere may be distinguished by its muted absorption bands. Irradiation from the
accretion shock also changes the chemical composition of the atmosphere, further altering spectra. Atmospheric
retrievals promise to map out the deposition of accretional energy, helping to decide between ”hot” and ”cold” start modes of
planet formation.
Read the full abstract ▾
All authors Nick Choksi, Peter Gao, Jerry Xuan, William Balmer, James Mang
Robotics
Junxiang Wang, Yuchen Liu, Yi Wu et al. (+6 more)
Memory allows large
language model (LLM)
agents to use past experience, often stored as textual records, when responding to requests. For
embodied agents that control
robots, visual and spatial observations introduce a broader choice of
memory modalities, affecting both the information available for subsequent tasks and the model inference costs of constructing and using
memory. We introduce MeMoREval, a
benchmark for studying this problem through a controlled experience-request protocol.
Agents first experience a sequence of events without knowing the subsequent tasks, and subsequently they use
memory to respond to information and action requests without receiving new observations. The
benchmark comprises 53 tasks across four simulated indoor environments and measures task success alongside the inference costs of
memory construction and use. Our reference evaluations reveal that storing frames outperforms storing generated descriptions at lower write-time inference cost for the primary model, while storing frames with scene graphs achieves the highest aggregate task success for every tested model. Extensions to accumulated experience, selective frame retention, and visual input to
memory writers demonstrate how the
benchmark accommodates further
memory-design investigations. Two physical-
robot demonstrations establish the feasibility of applying its protocol beyond simulation.
Read the full abstract ▾
All authors Junxiang Wang, Yuchen Liu, Yi Wu, Guofei Chen, Nir Pechuk, Tao Chen, Zackory Erickson, Yuzhe Qin, Tianyu Li
High Energy Physics (Theory)
Quantum limit to transparency
Angelo Esposito, Alberto Nicolis and Federico Piazza
We argue that no material can be made completely transparent to light as long as it has a nontrivial index of refraction. The
quantum decay of
photons into
phonons is an ineliminable source of attenuation. This survives even once all other excitations have been gapped away,
impurities have been removed, and the temperature has been set to zero. Restricting for simplicity to the case of a superfluid, we determine via effective field theory methods the corresponding low-frequency contribution to the imaginary part of the index of refraction. This scales as $\omega^4$ and induces a nonanalyticity of the real part for $\omega \rightarrow 0$. We also extend our result to large coherent electromagnetic fields, such as those of a
laser, showing that the effect can receive strong nonlinear enhancements. Our result is a genuine
quantum lower bound to the transparency of a medium.
Read the full abstract ▾
General Relativity and Quantum Cosmology
Asymptotic AEther Misalignment and Universal-Horizon Mechanics
Walter Arata, Moustafa Ismail, Stefano Liberati et al. (+3 more)
Universal horizons trap even arbitrarily fast signals, providing a test of black-hole thermodynamics in theories with a preferred time direction. In Einstein–aether theory, this direction is defined by a dynamical field, the aether, that can remain non-aligned relative to stationary time at infinity. In particular, for asymptotically anti-de Sitter black holes, it was so far impossible to recover a first law of black-hole mechanics when such asymptotic tilt of the aether is present. Here, we resolve this issue for an exact family of static, spherically symmetric solutions. Using the covariant phase-space formalism, we identify the asymptotic tilt and the decaying part of the aether field as a boundary source and its conjugate response. Allowing both to vary independently obstructs the definition of an energy that generates time translations. Fixing either quantity, or imposing a relation between them, restores such an energy, while externally varying the boundary data introduces work into the first law. The energies obtained by fixing the source or its response are related by a Legendre transformation. At the universal horizon, heat includes both the metric area contribution and an irreducible aether energy flux. The resulting balance separates work at infinity from horizon heat and explains why the physical temperature cannot generally be inferred from the coefficient of an area variation in the black-hole mass.
Read the full abstract ▾
All authors Walter Arata, Moustafa Ismail, Stefano Liberati, David Mattingly, Luke Martin, Giulio Neri
Robotics
PanTact: Teaching Whole-Body Dexterity with a Wearable Contact Interface
Xiaomeng Xu, Han Zhang, Zhian Ruan et al. (+6 more)
Humanoid robots have many degrees of freedom and contact surfaces, yet most loco-manipulation research restricts manipulation to the hands and avoids other body contacts with the environment. To extend manipulation beyond the hands, we present PanTact, a system that learns
humanoid whole-body dexterity, the capability to manipulate with any suitable body surface, directly from
robot-free human demonstrations. Our key idea is to use contact as manipulation-centric information that transfers across embodiments. Contact locations and timing specify the interactions needed to accomplish a task, while the
robot chooses how to coordinate its body to realize them within its own physical constraints. Specifically, PanTact captures, learns, and executes whole-body contact via: 1) a wearable,
robot-free data collection interface that augments
egocentric vision and bimanual UMI grippers with whole-body modular contact patches to record body contact locations and states; 2) a visuomotor policy that mitigates the
egocentric observation gap and predicts whole-body contact goals; and 3) contact-aware whole-body control formulations that realize active contact goals using suitable
robot surfaces. We deploy PanTact on a Unitree G1
humanoid without additional body contact sensors, demonstrating whole-body dexterity involving body support, pushing, sweeping, and load-sharing. Website: https://pantact.github.io
Read the full abstract ▾
All authors Xiaomeng Xu, Han Zhang, Zhian Ruan, Xiaofeng Guo, Jinkun Cao, Ruihan Yang, Carmelo Sferrazza, Shuran Song, Guanya Shi
Mesoscale and Nanoscale Physics
Frustrated Topological Orbital Magnetism
Sergio Garrido Bourkhaeva, Claudio Castelnovo and Wojciech J. Jankowski
By studying the interplay of
geometric frustration with Aharonov-Bohm fluxes in three dimensions, we unveil a rich playground exhibiting topological semimetallic ground states with orbital
magnetic dipole, quadrupole and ferrotoroidic orders, akin to
spin patterns in classical
spin ices - a class of phenomena which we dub frustrated topological orbital magnetism. We demonstrate that the ground-state toroidizations underpinned by non-Abelian topological nodal lines induce measurable current-induced
Hall phenomena. We further find that distinct 2-in-2-out and 3-in-1-out frustrated orbital current patterns can be associated with topological Weyl node charges, and corresponding anisotropic linear and nonlinear
Hall effects. We propose signatures for experimental validation of distinct frustrated topological orbital
magnetic orders in candidate materials.
Read the full abstract ▾
Quantum Physics
TLS loss in superconducting resonators on epitaxial hBN thin films
J. Christiansen-Salameh, D.J. Temples, R. Linehan et al. (+21 more)
The integration of low-loss dielectrics into superconducting
circuits is of importance to the functionality of
quantum computers and sensors. In this work, we fabricate and measure niobium resonators atop few-layer, epitaxial hexagonal boron nitride (hBN) grown by Molecular Beam Epitaxy on a sapphire substrate and compare against a control counterpart without hBN. We conduct radiofrequency measurements which probe the responses to power, temperature, and temporal fluctuations of two-level systems. The resonators integrating hBN maintain low dielectric loss with a low-power internal loss tangent on the order of $10^{-6}$, and demonstrate hBN can be added to the conventional superconducting
qubit stack with minimal degradation of the quality factor. This research represents a first step towards a materials-driven strategy to eliminate
phonon-mediated correlated errors in solid-state
quantum systems, which leverages the thermal properties of 2D materials to prevent athermal
phonons excited in the device substrate from transmitting across the
superconductor-substrate interface. The same technique applied to
quantum sensors would maximize
phonon collection by blocking
phonons from entering passive metal films.
Read the full abstract ▾
All authors J. Christiansen-Salameh, D.J. Temples, R. Linehan, S. Sussman, Aloyse Maille, F. Abouzahr, L. Alexis, K. Anyang, D. Bowring, G. Bratrud, E. Celi, A.H. Colón Cesaní, R. Khatiwada, D. Molenaar, O. Peek, A. Rodriguez, O. Seidel, D. Schlom, D. Sotir, S. Uemura, Y. Wu, E. Figueroa-Feliciano, D. Baxter, Z. Tian
Quantum Gases
Flat-band quantum geometry from embedded quench dynamics
Ilia Komissarov, Daniel E. Sheehy and Ahana Chakraborty
Quantum geometry underlies a wide range of correlated phenomena in electronic systems, yet accessing the geometry of individual bands remains experimentally challenging. Advances in
quantum simulators, such as ultracold atoms in optical
lattices, have enabled local state preparation and site-resolved measurements, offering direct access to
quantum dynamics beyond conventional transport probes. These developments naturally raise the question of whether band-resolved
quantum geometry can be extracted from local occupation readouts. Here, we propose an on-site quench protocol to probe the flat-band
quantum geometry via local occupation measurements in an atomic
lattice. Our scheme dynamically isolates the flat-band sector by coupling a one-dimensional flat-band
lattice to a two-dimensional flat-band bath. The bath selectively removes dispersive excitations, while compact localized states remain trapped due to their overlap with the flat-band modes of the bath. We show that the late-time occupation pattern directly encodes the integrated overlaps between flat-band states separated by the wavevector of the initial density pattern, thereby enabling access to integrated
quantum distances between the states separated by the commensurate momenta $\delta k = 2 \pi s/r$, $s,r \in \mathbb Z$ and to the integrated flat-band
quantum metric. Our results thereby establish a direct connection between
quantum geometry and the long-time
memory of an initial spatial pattern in flat-band systems.
Read the full abstract ▾
High Energy Physics (Theory)
GLSMs for non-polynomial deformations and generalized complete intersection Calabi-Yau manifolds
Lara B. Anderson, James Gray, Eric Sharpe et al. (+1 more)
Gauged Linear Sigma Models (GLSMs) have long been a key tool in understanding Calabi-Yau compactifications of string theory. While descriptions of complete intersections of
polynomial defining equations have long been understood in this context, a full treatment of non-
polynomial complex structure deformations has been absent from the literature. Likewise, the somewhat more exotic generalized complete intersection Calabi-Yau construction has resisted a GLSM description. In this paper we resolve both of these omissions using completely standard GLSMs with
polynomial superpotentials. The non-
polynomial structure seen in the
geometric constructions only arises in the associated GLSM upon restriction to the
vacuum space of the theory. This feature enables the construction to avoid singularities in field space that could have arisen if non-
polynomial structure were incorporated directly into the GLSM superpotential.
Read the full abstract ▾
All authors Lara B. Anderson, James Gray, Eric Sharpe, Albert Haochen Zhang
Astrophysics of Galaxies
A Single Line-Emitter Identified in an Unbiased Search with JWST/NIRSpec: Multi-Object Coronagraphy for Slitless Spectroscopy
Mason S. Huberty, Claudia Scarlata, Matthew J. Hayes et al. (+13 more)
We report the detection of a previously unknown object with a single emission line within the Hubble Ultra Deep Field, using a new JWST NIRSpec technique called Multi-Object Coronagraphy for Slitless Spectroscopy (MOCSS) with the G140M/F100LP observing mode. The MOCSS technique involves opening almost all the shutters on the NIRSpec microshutter array, while simultaneously closing the shutters that lie on top of known low
redshift objects, facilitating a wide field slitless spectroscopic search without contamination from bright objects. This observational technique has enabled the detection of a strong emission line in four different dither positions and in two position angles, and thus cannot be artificial. This line is detected at 1.60 $\rm \mu m$ with a line flux of $[2.5\pm 0.3]\times10^{-18} \rm erg\ s^{-1} cm^{-2}$ and a FWHM of $13.2\pm2.0$Å, and does not correspond to any known source with some of the deepest HST and JWST imaging in existence. Based upon the absence of imaging detection, we place a lower limit on the observed frame equivalent width (EW) of 7620Å on this object, which strongly limits the number of possible lines that this line can be. Given the wavelength, an EW limit that is near the ceiling of expectation, and the lack of other lines in the bandpass, this line cannot be easily explained by [OII] at $z=3.29$, [OIII] at $z=2.20$, H$\alpha$ at $z=1.44$, and Ly$\alpha$ at $z=12.16$. We discuss possible AGN and transient phenomena that could explain the finding.
Read the full abstract ▾
All authors Mason S. Huberty, Claudia Scarlata, Matthew J. Hayes, Alaina Henry, Danielle A. Berg, Vieri Cammelli, Richard Ellis, Anne E. Jaskot, Yu-Heng Lin, Vihang Mehta, Matilde Mingozzi, Pierluigi Monaco, Alberto Saldana-Lopez, Aayush Saxena, Jonathan C. Tan, Alice Young
Cosmology and Nongalactic Astrophysics
A Burst of Reionization in the Dark Ages
Neal Weiner, Soubhik Kumar, Wenzer Qin et al. (+3 more)
Primordial overdensities at small scales can be significantly larger than those observed at large scales. As a result, they can collapse much earlier, at
redshifts $z \sim 100$. In these collapsed
halos,
stars and black holes can form, where they can then contribute to an early reionization of the universe. If this epoch is temporary, the free electrons will recombine yielding a low level of ionization leading into
cosmic dawn. This contributes to the optical depth of the universe, and adds more weakly to low-$\ell$ polarization signals of ionization compared to later sources. We study the effects of a
redshift-localized source of reionization $\Delta \tau_{burst}$ in the context of a broad suite of cosmological data to constrain this scenario, finding a preference of approximately $\sim 2-2.5\sigma$ for a new contribution. Regardless of Planck reprocessing used (PR3 - SRoll2+plik or PR4 - LoLLiPoP+CamSpec/HiLLiPoP), the results are consistent, favoring a new contribution $\Delta \tau_{burst} \sim 0.012$ ($0.001 < \Delta \tau_{burst} < 0.022$ at 95%) and a similar total optical depth $\tau \sim 0.073$. The precise era of reionization is somewhat discrepant depending on the higher-$\ell$ reprocessing used, with HiLLiPoP favoring $z\sim 45$, CamSpec $z\sim 100$ and plik $z\sim 165$. If such a source is present at $40\lesssim z\lesssim150$, LiteBIRD should be able to detect it, while Simons Observatory should also constrain it.
Read the full abstract ▾
All authors Neal Weiner, Soubhik Kumar, Wenzer Qin, Zilu Zhou, Yacine Ali-Haïmoud, Hector Afonso G. Cruz
Quantum Physics
Engineering Ergodicity-Breaking and Stable Emergent Conservation Laws in a Generic Quantum Many-body System Under Strong Periodic Drive
Sagnik Chaudhuri, Anirban Das, Asmi Haldar et al. (+1 more)
A periodically driven
quantum many-body system is believed to approach a locally infinite-temperature-like state as dictated by the ergodicity hypothesis, when all the couplings/parameters are of the same order. When the drive is stronger than a threshold value, a large class of systems is shown to break this mandate, exhibiting Dynamical Freezing (DF): ergodicity is broken due to the emergence of local operators with permanent
memory of their initial values (PRX 11, 021008, 2021). However, an exception to DF has recently been noted in a particular setting (PRX 16, 011041, 2026). Here we identify the condition for ergodicity-breaking in a system under strong drive, and show how to stabilize DF by exact, local engineering. We further show that the Floquet
Hamiltonian $H_{eff} = (i/T)\ln{[U(T)]}$ is non-local when ordered by
spin-flips, in a non-DF system, while it turns quasi-local under our simple, local DF engineering that keeps the
Hamiltonian $H(t)$ still non-integrable at each $t$. This perplexing non-universality is absent in
equilibrium, where non-integrability is believed to imply ergodicity and thermalization universally. Our work paves a way to engineering novel ergodicity-broken interacting Floquet phases of matter in various
quantum platforms, and also endows DF-assisted
quantum error mitigation and state stabilization techniques (see Nature 653, 1027 (2026)) with unprecedented versatility and flexibility.
Read the full abstract ▾
All authors Sagnik Chaudhuri, Anirban Das, Asmi Haldar, Arnab Das
High Energy Physics (Theory)
Proving the OSV conjecture with the gravitational path integral
Nikolay Bobev, Jan Boruch, Luca V. Iliesiu et al. (+1 more)
The OSV
conjecture relates the helicity supertrace of black holes in $4d$ $\mathcal{N}=2$ supergravity obtained through a Calabi–Yau compactification of Type IIA string theory to the modulus squared of the topological string partition function evaluated at the attractor point in
moduli space for a given set of
magnetic black hole charges and electric chemical potentials. We find Euclidean black hole saddles that compute this index to all orders in the 4d supergravity derivative expansion and calculate their action using equivariant localization. The gravitational on-shell action localizes entirely to the north and south poles of each black hole, with the two contributions determined by the holomorphic and anti-holomorphic topological string free energies, respectively, evaluated at the attractor values of the
moduli. For single-center black hole saddles, this provides a proof of the OSV
conjecture to all orders in the topological string coupling. The same localization formula also applies to all saddles that contribute to the helicity supertrace, such as multi-center black holes, which leads to a non-perturbative refinement of the
conjecture.
Read the full abstract ▾
All authors Nikolay Bobev, Jan Boruch, Luca V. Iliesiu, Dawid Maskalaniec
Robotics
Fly-by-Code: Embodied Coding Agents for Aerial Manipulation with Active Visual and Physical Feedback
Jaewoo Lee, Jeongyeon Seo, Sihyun Cho et al. (+10 more)
Embodied coding
agents construct
robot policies from natural-
language instructions by generating executable programs and refining them from execution feedback. This refinement assumes that execution feedback reveals whether the task succeeded and, after a failure, provides sufficient evidence to guide program revision. Aerial manipulation violates this assumption in two ways. Payload constraints restrict onboard camera coverage and leave task-critical outcomes outside the
robot's view (partial observability), while strong dynamic coupling and limited control margins make physical interactions fragile (limited interaction robustness) and leave failure causes unresolved. Re-examining the same feedback cannot recover observations that were never collected or separate failure causes that the available evidence does not distinguish. To acquire this missing evidence, we introduce Active Feedback (AF), an active "View and Probe" framework. When execution feedback leaves unresolved questions, the
agent synthesizes a separate diagnostic program that runs in simulation to inspect unseen outcomes from informative viewpoints (View) or test hypotheses about failure causes through targeted physical interactions (Probe). Across six simulated aerial manipulation tasks with 60 trials per condition, AF reduces false completion declarations from 25 to 5 and raises final task success from 31.7% to 75% compared with terminal feedback under the same coding-
agent configuration and task-execution limit. Using shared
robot APIs, task programs refined with AF in simulation run on a physical aerial manipulator without task-program source-code modification and complete multi-stage manipulation tasks in the real world. These results show that executable diagnosis fills gaps in execution feedback for
embodied coding
agents, improving both completion assessment and program refinement in aerial manipulation.
Read the full abstract ▾
All authors Jaewoo Lee, Jeongyeon Seo, Sihyun Cho, Gyeongrak Choe, Yutong Wang, Bavin Saravanan, Jia-Bin Huang, Furong Huang, Sebastian Scherer, Guanya Shi, H. Jin Kim, Seungjae Lee, Dongjae Lee
Differential Geometry
Triality rotation of connections and Dirac operators on SU(3)
Joao Baptista
We show how triality, the
symmetry permuting the three 8-dimensional representations of $\mathrm{Spin}(8)$, can also appear as a
symmetry of
geometric objects such as connections and Dirac operators. Take any Riemannian metric $g$ on the
manifold $M = \mathrm{SU}(3)$ and let $\nabla$ be a metric connection on $TM$. Using an explicit model of Clifford multiplication by deformed Okubo products, we combine spinorial lifts with isometries $TM \simeq S^+ \simeq S^-$ to show that $\nabla$ can be rotated into two new metric connections on $TM$. So there is a canonical $\mathbb{Z}_3$-orbit of metric connections determined by $(g, \nabla)$, with no other choices. The three connections coincide when $g$ is bi-invariant and $\nabla$ belongs to the Cartan-Schouten line, but in general they are different. The rotations of $\nabla$ preserve the norm of the
curvature, so they define a new $\mathbb{Z}_3$-
symmetry of the
moduli space of Yang-Mills connections on $TM$. They also preserve the Chern-Simons 3-form and adjoint Yang-Mills-
Higgs functionals. After choosing an orientation on $M$, the rotated connections define three self-adjoint, Dirac-type operators on the spinor bundle. We prove that these operators are natural, in the sense that they commute with the Kosmann derivatives of spinors along all invariant Killing vector fields that preserve $\nabla$. We also show that they are not isospectral, in general. This is verified by computing the simplest
eigenvalue for the Levi-Civita connection and observing that it splits among the three operators as the bi-invariant metric on $M$ is deformed to one with isometry group $(\mathrm{SU}(3) \times \mathrm{SU}(2) \times \mathrm{U}(1)) / \mathbb{Z}_6$. Finally, we outline how the constructions in the paper may extend to general 8-
manifolds with a $\mathrm{PSU}(3)$-structure.
Read the full abstract ▾
High Energy Physics (Theory)
Lattice Monte Carlo of $\lambda\phi^4$ on Euclidean de Sitter
Angelo Caravano
We report the first Monte Carlo simulation of
quantum field theory on Euclidean de Sitter space, $S^4$. Rather than discretizing space, we use a spectral
lattice based on spherical harmonics, keeping all
symmetries exact at finite cutoff. For a massless $\lambda\phi^4$ theory, the zero-mode distribution reproduces the Starobinsky-Yokoyama
equilibrium at weak coupling, and deviations closely follow the one-loop renormalization of $\lambda$. At strong bare coupling, the renormalized quartic coupling saturates near the ceiling of the one-loop renormalization-group flow. This saturation is the de Sitter counterpart of the triviality mechanism that bounds the
Higgs mass in flat space. Our results confirm nonperturbatively that the infrared divergences of light scalars in de Sitter are correctly resummed by the
stochastic formalism. More broadly, this work opens up de Sitter space to
lattice Monte Carlo methods.
Read the full abstract ▾
Quantum Physics
Exact solutions of non-Markovian dissipative Ising models
Mohammad Maghrebi
Applications of
quantum devices and simulators require an accurate understanding of their dynamics as well as the effects of environmental noise. While some sources of noise can be modeled as Markovian dissipation, coupling to resonators or vibrational modes often produces non-Markovian dynamics as well, making it difficult to characterize the system, particularly for interacting many-body systems. Here, we identify a broad class of exactly solvable many-body
quantum systems with both non-Markovian dynamics and Markovian dissipation. We consider a prototypical model of
spins coupled to lossy bosonic modes that mediate Ising-type interactions. We exclude a transverse field but allow local
spin dissipation that need not commute with the interactions, including dephasing, relaxation, pumping, and even processes that generate coherences. Using an exact
stochastic representation, we decouple the many-body dynamics into the
evolution of individual
spins, yielding exact
spin observables and correlations without adiabatically eliminating the
bosons. In the Markovian limit, our solutions describe Ising models with correlated dephasing and nonreciprocal interactions. An example of
spin squeezing shows that both the achievable squeezing and its optimal preparation time depend strongly on the interplay between non-Markovian dynamics and Markovian dissipation.
Read the full abstract ▾
Computation and Language
AutoIndexer: Flexible-Order Decoding by Training Causal Models on Chains of Edits
Shu Ishida, Aliasghar Khani, Tianyu Zhang et al. (+3 more)
Autoregressive decoding is less suited to non-causally ordered data, e.g., cross-references or future dependencies, since their outputs cannot be revised post hoc, unlike iterative human
reasoning.
Diffusion LLMs partially mitigate this limitation but still cannot flexibly insert or remove arbitrary numbers of
tokens from a pre-generated context. We design AutoIndexer as a generative model with a novel training and decoding mechanism that simulates a Chain of Edits, allowing the model to flexibly modify its prior outputs in hindsight via
token insertions, substitutions, and deletions. The model learns to predict marker
tokens for opening and closing edits, simultaneously moving a cursor to positions in the existing
token sequence, analogous to a text editor. Training sequences are obtained by perturbing ground-truth labels with inverse chained edit operations, and the model learns to edit by undoing those inverses. The perturbation training allows AutoIndexer to be robust to erroneous context and adaptable to editing tasks such as code correction. We demonstrate that our method exhibits post-generation context-editing capabilities that most autoregressive and
diffusion models lack without compromising performance on standard LLM
benchmarks. The code is available at https://github.com/AutodeskAILab/autoindexer.
Read the full abstract ▾
All authors Shu Ishida, Aliasghar Khani, Tianyu Zhang, The Cong Luong, James Seale Smith, Adam Gaier
Machine Learning
Transformer-Based Masked Diffusion Models Provably Learn Distributions with Latent Structure
Jingda Wu and Changxiao Cai
Masked
diffusion models (MDMs) generate discrete data by iteratively unmasking multiple
tokens in parallel using a learned mask predictor. Their output carries two errors: parallelization error due to parallel sampling, and estimation error from the mask predictor learned with finite samples. Existing theory treats these errors separately, and the available statistical bounds typically scale with the number $V^L$ of length-$L$ sequences over a vocabulary of size $V$, suffering from the curse of dimensionality. In this work, we establish end-to-end statistical guarantees showing that
transformer-based MDMs learn structured distributions with a sample complexity governed by their latent structure. Specifically, we model this structure through mixtures of hidden Markov trees, which capture latent dependencies within components and heterogeneity across components. When the data distribution is well approximated by such a mixture, a
transformer-based MDM with parallel unmasking schedules achieves $\varepsilon$-accurate sampling in KL divergence with $\widetilde{O}(LK_{\mathrm{par}}/\varepsilon)$ training samples, where $K_{\mathrm{par}}$ is
polynomial in $L$, $V$, the number of hidden states, and the number of mixture components. The
transformer network and sampling procedure require no knowledge of the tree
topologies or component parameters, thereby adapting to the unknown latent structure automatically. Overall, our results provide theoretical justification for the empirical effectiveness of MDMs on real-world data.
Read the full abstract ▾
Number Theory
On $m$-fullness in number fields
Michiel Horikx and Marta Pieropan
We prove asymptotic formulas for the number of $m$-full ideals of bounded norm in a given ideal class and for the number of $m$-full algebraic integers of bounded Weil height.
Artificial Intelligence
Equal-Budget Monte Carlo Tree Search versus Randomized-Restart List Scheduling for DAG Workflows: A Controlled Equal-Budget Computational Study
Yuyu Zhang
Monte Carlo Tree Search (MCTS) is often proposed for scheduling directed-acyclic-graph (DAG) workflows, yet most studies never run the comparison that could refute it - giving a much simpler method the same search effort. We run that comparison under a controlled protocol. All arms receive an identical budget - the same number of candidate schedules built and evaluated: an MCTS scheduling method, preregistered as the main arm; a randomized-restart list scheduler with no search tree as the control; and a textbook-UCT MCTS arm as a registered follow-up on the same domain. The analysis plan was locked before data collection; an independent checker confirmed all 768 confirmation-stage schedules valid with correct finish times. At the main budget of 2048 evaluated schedules, MCTS is 2.88% worse than the restarts (95% CI [-3.62%,-2.15%]; the 72 of 96 held-out instances with positive baseline-relative headroom). The deficit is not budget starvation: on the separate exploration pool restarts self-improve by 1.85% where MCTS gains 0.24%, and 700 of 720 of MCTS's first results are merely the warm start it was handed. The studied selection rule locks deterministically onto the first-expanded child, keeping its trials correlated while restarts draw gains from the tails; the causal link to makespan remains an inference. A structural map locates where the tree stops paying: restarts win on sparse irregular workflows, no difference is detectable on regular layered ones, and dense ones leave no room to improve. Swapping the rule for textbook UCT removes most of the deficit (-0.26%, 95% CI [-0.71%,+0.23%]) without reversing its sign. Results are confined to a controlled synthetic domain (N in {8,16,32} tasks, 2-8 processors, budgets up to 2048); we make no claim about larger budgets.
Read the full abstract ▾
Machine Learning
Enriched Holomorphic Neural Networks for 2D and 3D Laplace Problems
Kalp Shah, Andrei Vlad Nicula, Mihnea Andrei Bârsan et al. (+4 more)
Boundary-value problems for the Laplace equation develop singular behaviour near corners, edges and junctions that smooth approximators resolve poorly. We combine holomorphic neural networks with singular functions derived from the local geometry and boundary conditions. The representation is harmonic by construction, so no interior residual is needed: the network fits the smooth remainder from boundary data alone, and the enrichment terms capture the singular behaviour. Singular exponents are trainable, so the construction also applies when the local expansion is not known in advance. We extend the approach to three dimensions using independent networks evaluated along complex null directions. The enriched model is the most accurate tested method on every problem. Its margin is several orders of magnitude where the enrichment captures the singularity exactly, and a factor of about 1.7-2.1 over the next-best method on a CdZnTe radiation detector with many singular sites, in both two and three dimensions.
Read the full abstract ▾
All authors Kalp Shah, Andrei Vlad Nicula, Mihnea Andrei Bârsan, Robert-Stefan Sofroni, Agneya Bhardwaj, Jan Jelínek, Alejandro Valverde Mahou
Cryptography and Security
Organizational Context for the Enterprise Security Brain
Eden Yavin, Ariel Zadok, Gal Engelberg et al. (+2 more)
Agents that investigate and decide on enterprise security reason over a machine-readable view of the environment: its assets, identities, configurations, relations, and findings. The correct judgment about an asset also depends on what the organization knows about it: who owns it, what business function it serves, which policy governs it, and which risks were accepted. Part of this knowledge can be inferred from machine state, such as an ownership tag; most of it is recorded by people in policies, audit reports, architecture diagrams, tickets, and conversations, outside machine state. We call this distance the organizational context gap. Closing it poses four challenges: extracting the right claims from heterogeneous organizational data and, at inference time, detecting claims made stale by live state, checking live state against the governing policy, and binding architecture diagrams, which carry the business logic, to live resources. We introduce organizational context as a layer of the enterprise security
brain and define it with a taxonomy and six properties its representation must preserve. We then define a context model in which a claim carries one meaning, an attestation records each source that asserts it, strength is derived from live evidence at read time, and claims resolve onto the entities of the asset graph. On a production environment spanning AWS, Okta, Google Workspace, and GitHub, extraction recovers 0.80 of a pooled 1,355-claim ground truth at 0.87 precision, with 0.5% of extracted claims unsupported by their source. An
agent reasoning over live state and resolved context raises the conflict in 0.97 of stale-finding tasks, and reaches answer correctness of 0.83 on policy-
alignment tasks and 0.86 on architecture tasks.
Read the full abstract ▾
All authors Eden Yavin, Ariel Zadok, Gal Engelberg, Elad Elouz, Leon Goldberg
Sound
Input Representation, Not Model Size or Pretraining Scale, Shapes Speaker Verification Robustness to Temporal Degradation
Erdem Baha Topbas, Srikanth Madikeri, Elisa Pellegrino et al. (+1 more)
Speech remains intelligible to listeners under heavy temporal degradation, but whether speaker identity survives is largely unexplored. We evaluate six automatic speaker verification (ASV) models, trained on temporally intact
speech, and tested under three degradations that remove different combinations of temporal cues. The models form three pairs, one per architecture family (convolutional,
transformer, conformer), each pairing a waveform-input with a mel filterbank-input model. All models degrade, but robustness tracks input representation rather than pretraining scale or model size. In all nine family-by-condition comparisons the filterbank-input model degrades less, with eight of nine 95% confidence intervals excluding zero, even when it is smaller or has less pretraining than its waveform counterpart. In the controlled convolutional pair, where the models differ only in input representation, the filterbank-input model's equal error rate rises from 1.1% to 17.1% under the heaviest tested degradation, against 2.9% to29.7% for its waveform counterpart.
Read the full abstract ▾
All authors Erdem Baha Topbas, Srikanth Madikeri, Elisa Pellegrino, Volker Dellwo