arXiv 2611.14516 · cs.LG · stat.ML · Nov 10
TorchCurves: Differentiable Parametric Curves
Alex Shtoff
We present TorchCurves, a PyTorch library that treats vector-valued parametric curves as learnable primitives. Its central operation maps batches of scalar inputs to points on curves represented in B-spline or Legendre polynomial bases, with differentiation through both inputs and coefficients. Curve coefficients can be optimized jointly with the surrounding model or produced by another network. T
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14515 · cs.LG · cs.CL · Nov 10
AutoScaffold: AI-Guided Scaffolding for Reinforcement Learning of Language Models
Zijian Zhang et al.
Reinforcement learning with verifiable rewards (RLVR) produces rich training experience, yet outcome rewards reduce each attempt to a single number and leave most of this information unused. Training-time scaffolds, such as strategies or hints added to the prompt during training, offer a natural way to turn this information into guidance, but their design and adaptation are largely left to humans
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14514 · cs.AI · cs.CL · cs.LG · Nov 10
Contrastive Discovery: Open-Ended Scientific Discovery over Competing Explanations
Ziang Liu et al.
LLM agents are increasingly used for open-ended, data-driven scientific discovery, yet most of them evaluate hypotheses pointwise, testing an empirical claim $h$ only against its complement $\neg h$. When evidence rejects $h$, the agent learns that one explanation is unlikely but not which alternative the evidence supports. To recover that alternative, we introduce Contrastive Discovery, which rep
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14204 · cs.DC · cs.PF · cs.PL · Nov 10
Performance Portability Explained: Insights from accelerating ICON microphysics
Faveo Hörold et al.
Scientific codes are under pressure to run on HPC machines shaped by AI workloads, without a rewrite every time the hardware moves. This paper looks at accelerating ICON microphysics and argues that starting from the scientific problem — not a double-precision reference binary — opens more room for performance-portable implementations.
Why it matters
Domain scientists shouldn’t have to rewrite a microphysics scheme every time the machine is shaped by someone else’s AI workload. Portability that starts from the science, not a reference binary, is the practical way out.
arXiv 2611.14201 · cs.PL · Nov 10
CFGKAT: A Complete Calculus for Non-Local Control Flow
Spencer Van Koevering et al.
Guarded Kleene algebra with tests can reason about programs with if and while, and it has a fast decision procedure. Real languages also have break and goto. CFGKAT adds that non-local control flow, and this paper gives it a sound, complete axiomatization of program equivalence — something the extra jumps made surprisingly hard.
Why it matters
Compilers and verifiers need equality proofs that survive goto and break, not just structured loops. A complete calculus is how you stop treating those jumps as unreasoned special cases.
arXiv 2611.14147 · quant-ph · cs.CR · cs.PL · Nov 10
Secure Information Flow in Quantum Programs
Zhicheng Zhang et al.
Information-flow security tries to stop confidential data leaking through a program’s behavior. In quantum programs the state can be superposed or entangled, and execution mixes classical and quantum variables. This paper builds semantic foundations and verification methods for those leaks.
Why it matters
Quantum programs can leak through entanglement, not just through a printed value. Information-flow rules that understand superposition are how you reason about secrecy in that setting.
arXiv 2611.14488 · cs.RO · Nov 10
Is Your Robot's Memory Just a Clock? Temporal Shortcuts in Memory-Augmented VLA Policies
Chubin Zhang and Bo An
Memory-augmented robot policies are supposed to remember what happened in an episode. Many of them just sample frames on a clock and stamp each one with its time step. This paper shows that a brief pause or a fidget — something that changes nothing about the task — is enough to throw those memories off, because the policy has learned to treat time passing as if it were work.
Why it matters
If a robot’s “memory” is mostly a clock, waiting looks like doing. That’s a dangerous shortcut for any policy you would trust around people or parts.
arXiv 2611.14470 · cs.RO · cs.CV · Nov 10
TanCRAFT: Context-Reasoned Anticipatory Feedback for Assembly Tasks
Sassan Mokhtar et al.
Preventing human error in physical assembly tasks requires assistance systems that can intervene before mistakes are made. This is especially challenging in settings where users are free to choose the construction order and correctness depends on the evolving assembly rather than a fixed next step. We present TanCRAFT, a four-stage real-time predictive assistance system that tracks the scene state
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14440 · cs.CV · cs.LG · cs.RO · Nov 10
UGG: Unified Generative Geometry Model
Amani Kiruga et al.
Understanding 4D scenes is a core challenge in visual intelligence, requiring models to jointly capture scene geometry, appearance, and dynamics from partial observations. However, existing approaches are subdivided into feedforward models for direct geometric prediction, and generative models for visual synthesis. We introduce UGG, a unified framework for 4D scene understanding that jointly model
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14485 · math.NA · Nov 10
A fourth-order dimensionally split exponential time differencing scheme with real and distinct pole rational approximation for non-linear reaction-diffusion systems
Wisdom Attipoe and Emmanuel Asante-Asamani
This paper presents a novel fourth-order, L-stable exponential time differencing (ETD) Runge-Kutta type scheme incorporating a dimensional splitting integrating factor method for the numerical solution of nonlinear reaction-diffusion systems. The method, ETDRK4RDP-IF, is obtained by approximating the matrix exponential operators in the classical dimensionally split ETDRK-IF scheme with an accurate
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14349 · math.NA · Nov 10
Shape Perturbation of the Galerkin Boundary Element Dirichlet-to-Neumann Matrix on Moving Polygons
Tokuhiro Eto
For nonlocal interface evolution problems such as the Mullins-Sekerka flow, certain parametric schemes discretize only the interface and approximate the Dirichlet-to-Neumann map by boundary elements. Their error analysis requires bounds on how this discrete operator changes as the polygon moves. In that analysis, the polygons are only Lipschitz and close to a smooth curve only in a Lipschitz norm,
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14426 · math.OC · cs.AI · Nov 10
Two Communities, One Problem: Branch-and-Bound vs. A* for Resource-Constrained Project Scheduling
Izack Cohen et al.
Branch-and-bound comes from operations research; A* comes from AI. They share a search skeleton but explore the space differently, and that can change how long a solve takes. This paper lines them up on the resource-constrained project scheduling problem, spells out how their ingredients correspond, and measures where each one pulls ahead.
Why it matters
Scheduling, compilers, and planners keep rediscovering the same search. A head-to-head on one canonical problem is how those communities stop talking past each other.
arXiv 2611.14423 · physics.bio-ph · math.OC · q-bio.PE · Nov 10
Cotangent-lift inference of stochastic microbial dynamics
José A. Capitán et al.
We study the parameter estimation for the stochastic logistic and Lotka--Volterra diffusions from discrete, irregularly spaced observations corrupted by measurement noise, a setting typical of microbiome time series. For the logistic equation, the growth rate acts as a time change and does not enter the stationary law, so it can be identified only from the temporal structure of the data. We prove
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14158 · physics.comp-ph · cs.CE · Nov 10
Neural network scaling and generalization for 3D strain fields in heterogeneous materials
Lucas Fourel et al.
Multiscale materials simulations keep solving mechanical fields on high-resolution 3D volumes, which gets expensive fast. Neural surrogates can take over some of that work, but most are trained for a fixed loading and don’t say how the network should grow when the volume gets finer. This paper studies how those strain-field networks scale and whether they still generalize.
Why it matters
If a surrogate only works at the resolution you trained, it isn’t a substitute for homogenization — it’s a one-off fit. Scaling laws are how you decide whether the shortcut will hold.
arXiv 2611.14119 · physics.optics · physics.comp-ph · Nov 10
Physics-informed Bayesian optimization of high-order harmonic generation in gases
Dominika Maslarova et al.
High-order harmonic generation turns an infrared laser into extreme-ultraviolet light, but the conversion efficiency depends on a messy tradeoff between the single-atom response and macroscopic phase matching. The useful region looks like a thin ridge in pressure and length. This paper uses physics-informed Bayesian optimization to find it.
Why it matters
When the landscape has a thin ridge, a generic search wastes shots. Building the known physics into the optimizer is how you spend experiments on the part of the space that matters.
arXiv 2611.14484 · quant-ph · Nov 10
Nonperturbative string dynamics in a (2+1)D lattice gauge theory on a superconducting quantum processor
Sanziana Stelea et al.
Electric flux strings underpin confinement in gauge theories. In two spatial dimensions, their evolution involves competing processes of charge motion, pair creation, and transverse deformation. Accessing these dynamics beyond perturbative limits challenges classical approaches, as entanglement growth restricts the times accessible to tensor-network simulations. Here we use quantum simulation to i
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14461 · quant-ph · Nov 10
Finite-Key Readiness and Authentication-Key Replenishment in QKD-Assisted UAV Edge Computing
Dang H. Le et al.
A service that needs fresh quantum keys must wait for key release, and may be denied it, before protected computation starts. We quantify this cost for a trusted stationary UAV serving two ground gateways, each obtaining its keys from one finite-key decoy-state BB84 attempt with parameters fixed in advance. A count-level Monte Carlo model, reproduced by a semi-analytic model, links key acceptance
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14482 · gr-qc · Nov 10
Singustars: Between Black Holes and Compact Stars
Pablo Laguna
The Tolman--Oppenheimer--Volkoff equations admit singular solutions in which a negative point mass is surrounded by a self-gravitating fluid. For a $Γ$-law polytropic equation of state, these solutions, which we call singustars, are thick shells bounded by two zero-pressure surfaces around a naked, repulsive, negative-mass singularity, with a Schwarzschild exterior. They form a two-parameter famil
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.
arXiv 2611.14455 · gr-qc · hep-th · math-ph · Nov 10
Gravitational radiation from a particle in circular orbit around a wormhole: the nonrotating case
Shauvik Biswas and Sayan Chakrabarti
In this paper, we have generalized Poisson's(1993) calculation for the case of gravitational radiation from a particle in a circular orbit around a black hole to the case of the wormhole. Considering the motion of a test particle around a Schwarzschild like wormhole and using Teukolsky formalism, we have obtained an analytical result of flux at infinity under the assumption of small velocities. Fo
Why it matters
A new paper on this week’s radar — worth opening if you work in this lane.