Cryptography and Security
Unmasking the Secret Code: Using Canaries to Show How Masked Context Controls Memorization in SFT
Avi Schwarzschild, Om Thakkar, Sahra Ghalebikesabi et al. (+2 more)
Canaries, or synthetic training records containing known secrets, are widely used to assess whether
language models reveal information from their training data. But how much do these measurements depend on the canaries themselves? We study this question through controlled fine-tuning experiments with
language and
vision-
language models, spanning synthetic tasks and tasks where training improves useful performance. Whether a secret is provided in the training input strongly changes what the model learns; for example, models can reproduce a secret in the input yet fail to recall it when it is removed, even after extensive repetition during training. When training instead requires predicting the secret from the surrounding context, later recall can become reliable. Repetition alone is therefore insufficient to characterize memorization. We also find that changing when a canary appears, or which other canaries accompany it, can substantially change measured memorization without increasing that canary's repetition count. These results show that canary measurements depend on how secrets are presented during training and tested afterward. Audits should test whether their conclusions hold across these choices before extending them to other training records.
Read the full abstract ▾
All authors Avi Schwarzschild, Om Thakkar, Sahra Ghalebikesabi, Congzheng Song, Vinnie Monaco
Computational Physics
Global-in-time Spectral Duhamel-Picard Method for Post-Exposure Bake Simulation and Physics-Informed Neural Operators for EUV Lithography
Doyun Kim, Mircea Dusa, Werner Gilijns et al. (+1 more)
We present a global-in-time spectral Duhamel-Picard method with
polynomial expansions for post-exposure bake (PEB) simulation. The coupled acid-quencher reaction-
diffusion equations are reformulated as nonlinear Volterra integral equations using the Duhamel principle. Full trajectories are represented by finite
polynomial expansions, with temporal coefficient fields determined over the bake interval by Anderson-accelerated Picard iteration, thereby avoiding sequential time marching. In 3-D PEB simulations, we employ Chebyshev and Legendre temporal bases while evaluating the Duhamel integrals using the corresponding Gaussian quadrature rules. We demonstrate that the proposed method simulates a large domain of $1200\times1200\times100$ voxels on the order of one minute. The representation is further incorporated into a physics-informed neural operator (PINO) with a mixed-transform variant of a factorized
Fourier neural operator, which is trained without precomputed solution trajectories. Together, the integral formulation and its PINO extension enable accelerated, time-marching-free prediction of concentration fields at arbitrary times within the bake interval.
Read the full abstract ▾
All authors Doyun Kim, Mircea Dusa, Werner Gilijns, Philippe Leray
Quantum Physics
The noise cost of reservoir locality in broadband quantum amplifier chains
Chengxi Li, Hao Zhu, Tie-Fu Zhang et al. (+3 more)
Directional
quantum amplifiers enable weak-signal readout while protecting the source, but reservoir noise limits sensitivity. Finite-band suppression must reconcile frequency-dependent responses with fixed local couplings and finite rates. We construct adjacent-site loss channels attaining the known single-frequency minimum under local compatibility conditions, as an uncapped
benchmark. For specified uniform and dimerized chains, we compare fixed reservoirs at matched drift, ports, high-gain band and complete gain/loss rate caps. At the representative cap $R=8$ and halfwidth $\Omega=0.05$, nearest-neighbor arithmetic band average noise exceeds the unrestricted optimum by more than $1.01265$ quanta in the uniform chain for every finite $N\ge64$. Boundary-response bounds give necessary rates and targets requiring range two. At $N=64$, finite-auxiliary calculations retain a noise advantage under matched total coupling budgets; response bounds quantify noise-resolution tolerances. These theoretical results guide reservoir design and noise calibration for a proposed cryogenic microwave array with a $100$ kHz design band.
Read the full abstract ▾
All authors Chengxi Li, Hao Zhu, Tie-Fu Zhang, Lin Zhuang, Wanzi Sun, Wuming Liu
Astrophysics of Galaxies
A JWST View of Cosmic Star Formation from $z\sim0$ to 9 with Hydrogen Recombination Lines in the COSMOS and GOODS-N fields
Shuqi Fu, Linhua Jiang, Christopher J. Conselice et al. (+9 more)
We combine wide-area COSMOS-3D observations with deeper GOODS-N data from JWST/NIRCam F444W wide-field slitless spectroscopy to trace
cosmic star formation over $z\sim0$–9. The observations cover approximately $0.33~\mathrm{deg}^2$ in COSMOS and $60.9~\mathrm{arcmin}^2$ in GOODS-N, with typical $3\sigma$ integrated line-flux limits of $3.2\times10^{-18}$ and $1.04\times10^{-18}~\mathrm{erg\,s^{-1}\,cm^{-2}}$, respectively. The common $3.9$–$5.0\,\mu$m wavelength window probes Br$\alpha$, Br$\beta$, Pa$\alpha$, Pa$\beta$, H$\alpha$, and H$\beta$ across the local Universe,
cosmic noon, and reionization. Our final sample contains 9146
galaxies, including 7883 in COSMOS and 1263 in GOODS-N. COSMOS constrains luminous emitters, while GOODS-N extends the luminosity functions toward fainter sources. We measure dust-corrected line luminosities, account for line-detection completeness, and convert the six lines to equivalent H$\alpha$ luminosities assuming Case B recombination. Schechter fits reveal an increase in the characteristic H$\alpha$-equivalent luminosity toward
cosmic noon. Integrating the resulting
star formation rate functions above $0.24\,M_\odot\,\mathrm{yr}^{-1}$, we trace the
cosmic star formation rate density from the local Universe to reionization. The combined Brackett, Paschen, and Balmer measurements are broadly consistent with the canonical UV- and IR-based
cosmic star formation history at $z\lesssim6$. Fixed-slope tests show that the H$\beta$ estimate at $z\sim8$ remains sensitive to the external photometric faint-end constraint and extrapolation below our spectroscopic limit. These measurements provide complementary constraints on the distribution and integrated rate of
star formation across
cosmic time.
Read the full abstract ▾
All authors Shuqi Fu, Linhua Jiang, Christopher J. Conselice, Zijian Zhang, Danyang Jiang, Jie Chen, Louis Quilley, Fengwu Sun, Ronaldo Laishram, Alberto Traina, Bahareh S. Salmasi, Duncan Austin
Astrophysics of Galaxies
Deep X-ray View of a Distant Protocluster Anchored by a Quasar at $z=6.63$: No Evidence for Enhanced X-ray AGN Activity
Shouyi Wang, Feige Wang, Jinyi Yang et al. (+22 more)
Whether dense environments accelerate SMBH growth during
cosmic reionization remains uncertain because X-ray constraints on $z>6$ protoclusters are rare. We analyze $\sim865$ ks of Chandra observations of J0910-0414
quasar field, a massive protocluster at $z=6.63$. We construct a catalog of 251 X-ray sources and search for AGN among 21 spectroscopically confirmed members, excluding the central
quasar. This is the first systematic X-ray AGN census in a $z>6$ protocluster; none of the members is individually detected. Stacking all members ($\sim17$ Ms total exposure) yields no significant signal and limits the average rest-frame 2-10 keV luminosity to $\langle L_{\rm 2-10~keV}\rangle\leq9.54\times10^{42}~\rm erg~s^{-1}$. The null result gives a central 90% Jeffreys upper bound of $f_{\rm AGN}<0.09$. After accounting for the spatially varying sensitivity, the field Compton-thin X-ray luminosity function predicts $N_{\rm exp,field}=0.0356$ AGN in the surveyed volume. Assuming the overdensity profile of AGNs in this protocluster is the same as that of
galaxies determined by Wang et al. (2024), the expected AGN count becomes $N_{\rm exp}=0.131$. Zero detections then imply a one-sided 90% upper limit of $\eta_{\rm rel}<17.5$ on any additional X-ray AGN enhancement beyond that associated with the
galaxy overdensity. Although $\eta_{\rm rel}<17.5$ permits more than an order-of-magnitude enhancement, the data show no significant excess of X-ray-detectable AGN in J0910-0414 protocluster. The present X-ray data may miss heavily obscured, intrinsically X-ray-weak, or intermittently accreting SMBHs; the non-detection therefore does not rule out AGNs among the member
galaxies.
Read the full abstract ▾
All authors Shouyi Wang, Feige Wang, Jinyi Yang, Bin Luo, Xiangyu Jin, Fan Zou, Jaclyn B. Champagne, Xiaohui Fan, Elaina Malczewski, Fabio Vito, Xue-Bing Wu, Eduardo Bañados, Fuyan Bian, Thomas Connor, Tiago Costa, Anna-Christina Eilers, Emanuele Paolo Farina, Joseph F. Hennawi, Hyunsung D. Jun, Jiang-Tao Li, Chiara Mazzucchelli, Klaudia Protušová, Fabian Walter, Zhibo Yu, Minghao Yue
Game Theory
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows
Jiayuan Liu, Mingyu Guo, Jiarui Gan et al. (+1 more)
Assigning routine tasks to cheaper large
language model (LLM) providers can reduce generation cost, but switching providers may require processing accumulated context again. Keeping dependent tasks with the same provider can instead reuse eligible context. Graph-based workflows therefore couple routing decisions across tasks. Independent providers may also privately know their real-time unit costs and strategically misreport them. We formulate this routing problem as a single-parameter procurement auction and propose
STAR (Strategy-proof
Topology-aware
Agent Routing), which minimizes virtual-cost-weighted graph workloads. Its allocation rule and payments, computed by integrating workload over possible cost reports, make truthful reporting utility-maximizing regardless of other providers' reports and cover each truthful provider's execution cost. With independent cost distributions and non-decreasing virtual costs,
STAR additionally minimizes the user's expected total payment to providers. Controlled synthetic experiments separate the execution-cost benefit of
topology awareness from the payment benefit of virtual-cost optimization: the former grows with fan-out and transferred payload, while the latter increases gradually along the tested prior-shape heterogeneity path. A paired sensitivity study quantifies how payment savings vary with prior misspecification while strategy-proofness is preserved.
Read the full abstract ▾
All authors Jiayuan Liu, Mingyu Guo, Jiarui Gan, Vincent Conitzer
Machine Learning
Historical-Feature Residual Adaptation for Sequential Screening
Yixiang Hong and Yuchen Zhang
Sequential screening policies must use acquired feedback effectively under a fixed experiment budget. We study acquisition-time residual adaptation: a small ranking correction built from fixed historical features and anchored to base scores recorded before feedback is revealed. The interface accepts a history-conditioned scorer and a candidate feature table; its primary instantiation augments AssayFormer's ten-dimensional representation with 64 historical directions while retaining the pretrained model. In retrospective evaluation on 20 official-test screens from 15 recorded sources, a comparison fixed before target-outcome access increased source-macro fraction of hits recovered from 22.58% to 28.61%, exceeding all six prespecified comparators. A same-feature ordinary ranking head achieved 25.73%. Supplementary strong logistic, rank-fusion, and unprojected controls achieved 26.42%, 26.02%, and 26.78%, respectively. A separate model-independent pilot on 13 reused development screens reconstructs the features without AssayFormer assets. Across three seeds, residual adaptation improves random-forest hit recovery from 28.87% to 31.27%, while logistic recovery changes from 30.22% to 29.78%. The results show effective pretrained-policy adaptation and useful transfer to an independent nonlinear base, with gains that depend on the base model.
Read the full abstract ▾
Computation and Language
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information?
Xinyi Liu, Hooshang Nayyeri, Emine Yilmaz et al. (+3 more)
Large
language model (LLM) evaluators are increasingly used to assess personalized assistants, yet whether they can reliably judge personalization has rarely been tested. We introduce PersEvalBench, a
benchmark testing whether an evaluator can recognize appropriate and inappropriate uses of user history. It contains 403 multi-episode scenarios where user information should persist, transfer, or cease to apply. The
benchmark covers nine controlled error subtypes across three families: under-personalization, over-personalization, and misapplied-state errors. Human audits confirm that generated errors match their intended types and that the prescribed uses of user information are reasonable. Across five LLM evaluators, recognition of controlled personalization errors remains limited even with the full dialogue history. On the human-evaluated subset, under-personalization shows the largest gap: humans recognize 94.4% of these errors, compared with 63.9% for the strongest reported evaluator. Evaluators often recognize that a response is problematic but reverse the error direction, mistaking failure to follow an applicable preference (UNDER) for use of an inapplicable preference (OVER). Easier evidence access offers little help: highlighting the relevant history turns raises error recognition only slightly, suggesting that the remaining difficulty lies in judging what role that information should play in the response.
Read the full abstract ▾
All authors Xinyi Liu, Hooshang Nayyeri, Emine Yilmaz, Dilek Hakkani-Tur, Charith Peris, Hari Thadakamalla
Solar and Stellar Astrophysics
Transients and Stellar Activity as Revealed by Ultraviolet Diagnostic Tracers
Richard de Grijs, Martin A. Barstow, Deepak Chahal et al. (+5 more)
Ultraviolet (UV) observations offer unique insights into transient and variable astrophysical phenomena, from the Sun and nearby
stars to
supernovae and
star-forming
galaxies. Here, we summarise and review the diagnostic power of the UV spectrum for probing the physical conditions, dynamics and energetics of these events. We trace the historical development of UV astronomy from early satellite missions to the current landscape shaped by the Hubble Space
Telescope, the
GALaxy Evolution eXplorer, the Neil Gehrels Swift Observatory and AstroSat/UVIT. Particular emphasis is placed on resonance lines such as Mg II $h$ and $k$, C II, Si IV, C IV and N V, which probe the (solar and) stellar chromosphere and transition region, revealing
magnetic heating, shocks and mass loss across stellar types. We examine UV diagnostics of stellar
flares, pre-main-sequence activity and rotation–activity correlations, linking them to exoplanetary irradiation and habitability. We then explore the UV view of
supernova shock breakout and early-time emission, as well as the far-UV mapping of young stellar populations and hierarchical
star formation in the Magellanic Clouds. Together, these aspects illustrate how the UV connects small-scale
plasma processes to galactic-scale
star formation. Despite remarkable progress, UV transient astronomy remains limited by instrument availability and cadence. Forthcoming missions promise to transform this field by capturing the rapid, luminous phenomena that define the dynamic Universe.
Read the full abstract ▾
All authors Richard de Grijs, Martin A. Barstow, Deepak Chahal, Fabio Favata, Xiaoting Fu, Ana I. Gómez de Castro, Sipra Hota, Firoza Sutaria
Astrophysics of Galaxies
The Contribution of Mergers and Accretion to Dark Matter Halo Assembly: An Excursion Set Theory Approach
Hamed Kameli
We study the relative contributions of mergers and
accretion to the assembly of dark matter
haloes within the framework of Excursion Set Theory (EST). We extend a variance-discretized merger-tree
algorithm, which recursively reconstructs all branches of a complete merger tree and introduces a merger-ratio criterion for distinguishing mergers from
accretion. The method is validated by comparing the resulting progenitor distributions with the analytical conditional first up-crossing distribution, showing excellent agreement over a wide range of
halo masses and
redshifts. We find that the
accretion contribution depends strongly on the adopted merger-ratio threshold. For large merger ratios ($MR=0.1$–$0.25$),
accretion accounts for up to $\sim70\%$ of the mass growth of massive
haloes, whereas including all resolved mergers ($MR=10^{-6}$) reduces the
accretion contribution to $\sim40\%$. These values are in good agreement with previous N-body simulation results. The accreted mass exhibits a nearly self-similar distribution in the mass-
redshift plane, with the dominant contribution arising from progenitors in the range $10^{-3}M_0 \lesssim M \lesssim 10^{-1}M_0$ and predominantly at $0.5 \lesssim z \lesssim 2$. Our results demonstrate that EST-based merger trees provide an accurate and computationally efficient framework for investigating
halo assembly and quantifying the relative roles of mergers and
accretion in hierarchical structure formation.
Read the full abstract ▾
Number Theory
Shapes of tame abelian number fields
Anwesh Ray
The shape of a number field $K$ of degree $n\geq2$ is the similarity class of the
lattice obtained by projecting $\mathscr{O}_K$ orthogonally away from the rational line in Minkowski space. It defines a point in the space of shapes of rank-$(n-1)$
lattices, $\mathscr{S}_{n-1}:=\mathrm{GL}_{n-1}(\mathbb{Z})\backslash \mathrm{GL}_{n-1}(\mathbb{R})/\mathrm{GO}_{n-1}(\mathbb{R})$, where $\mathrm{GO}_{n-1}(\mathbb{R})$ is the group of orthogonal similitudes. Let $G$ be a fixed nontrivial finite abelian group, and let $K/\mathbb{Q}$ vary over tamely ramified Galois extensions with Galois group isomorphic to $G$. We study the variation of the shapes of such number fields in $\mathscr S_{n-1}$, where $n=|G|$.
Read the full abstract ▾
Mathematical Physics
A Nontrivial Lattice Scalar Quantum Field Model
P.A. Faria da Veiga, R. Mistry and M. O'Carroll
We introduce a new Euclidean
quantum model with zero-
spin scaled scalar fields on a spacetime hypercubic
lattice, which we refer to as the hybrid model. The model coupling parameters are the hopping parameter $\kappa^2>0$ and the field interaction coupling $\alpha>0$. Using the technique of
lattice scaled fields, with a noncanonical scaling that corresponds to a partial renormalization (wave function renormalization), we show that the partition function of this model satisfies both thermodynamic and ultraviolet stability bounds, with constants which are uniform in the
lattice volume $\Lambda_s$ and the
lattice spacing $a\in(0,1]$. As a consequence, the model has a finite free energy per degree of freedom in the thermodynamic limit $\Lambda_s\nearrow\infty$, at least in the subsequential sense. Subsequently, we can also take the ultraviolet or continuum limit $a\searrow 0$. Besides the finiteness property of the free energy, for sufficiently small values of the model coupling parameters $\kappa^2$ and $\alpha$, we show the model satisfies Osterwalder-Schrader positivity and construct the underlying physical
quantum mechanical
Hilbert space, and, using a
polymer expansion, we show correlations also remain bounded in the thermodynamic limit. Finally, we perturbatively compute an infinite volume truncated four-point correlation showing it is nonzero. Noting that, for fixed $a$, perturbation series in $\kappa^2$ and $\alpha$ converges, these results are sufficient to show that our hybrid model makes sense as a
lattice quantum field theory in spacetime dimension $d=3,4$, with possible extensions to other higher values of $d$.
Read the full abstract ▾
Number Theory
Ternary polynomial relations and periods of cusp forms
Vasily Ilin and Ken Ono
Bachmann and van Ittersum conjectured that every
polynomial $f\in\mathbb{Q}[x,y,z]$ satisfying $$ f(x,y,z)=f(y,z,x),\qquad f(x,y,z)+f(-y,x-y,z)+f(y-x,-x,z)=0 $$ is a sum of
polynomials in two variables. We prove this
conjecture over every field of characteristic zero and determine all homogeneous solutions. Moreover, restriction to $z=0$ identifies the degree $n$ solution space with the space of binary period
polynomials of degree $n$ normalized to vanish on both coordinate axes. The solution corresponding to such a
polynomial $q$ is $$ f(x,y,z)=q(x,y)+q(y,z)+q(z,x). $$ There are no nonzero homogeneous solutions of odd degree. For even degrees $n$, the symmetric and alternating solution spaces are each isomorphic over $\mathbb{C}$ to the space of weight $n+2$ cusp forms on $SL_2(\mathbb{Z})$. The proof connects the
polynomial relations with the cohomology of arithmetic groups through the rational Tits building for $GL_3$. Cusp forms enter both the description of the solutions and the proof that there are no others. The main
theorems and the essential argument are formalized unconditionally in Lean~4 by AxiomProver.
Read the full abstract ▾
Distributed Computing
Parallel Sparse Outputs for Tensor Compilers
Adrian Gushin, Pau Sum and Willow Ahrens
Sparse tensor compilers are commonly used high-performance tools that optimize tensor
algebra programs over notoriously complex sparse data structures. Despite the ever-increasing size and scale of sparse
datasets, there exists astonishingly little research on developing a parallel sparse tensor compiler. Existing approaches fail to parallelize or make assumptions about access patterns or output formats that limit generality. This paper presents Birdseed, the first tensor compiler to support all loop orderings over any standard sparse format in parallel. More specifically, Birdseed supports parallel sparse scatters and directly writes into list, bytemap, hash, and dense levels of fibertree sparse formats. Our approach exhibits asymptotic
memory and time improvements over all previous work. We additionally demonstrate empirical speedups on challenging kernels to parallelize, such as histogram construction and SpMSpV.
Read the full abstract ▾
Number Theory
Zero-cycles and Brauer groups on certain Enriques surfaces
Thomas Jaklitsch
In this paper, we study the Chow group of zero-cycles $\text{CH}_0(S)$ for certain Enriques quotients $S$ of the Kummer surface associated to the product of two geometrically non-isogenous elliptic curves $E_1$ and $E_2$. We show that over fields of characteristic different from $2$ or $3$, the degree-zero subgroup $A_0(S) \subset \text{CH}_0(S)$ is 16-torsion. Over p-adic fields, we compute the transcendental Brauer group $\text{Br}(S)/\text{Br}_1(S)$. Using the Brauer-Manin pairing, we construct examples of Enriques surfaces for which $A_0(S)$ attains nontrivial 2-torsion elements arising from push-forward from the abelian surface $E_1\times E_2$.
Read the full abstract ▾
Machine Learning
Did the LLM Answer the Question? A Semantic Reconstruction Game for Jailbreak Evaluation
Liam Davis, Romain Emil-Popescu and Suman Jana
Jailbreak attacks against large
language models (LLMs) are evaluated using attack success rate (ASR), which depends on correctly identifying responses that fulfill prohibited requests. Existing evaluators can mistake refusals or topically related responses for successful attacks. We show that changing only the judge LLM can alter reported ASR by up to $82$ percentage points. We introduce the semantic reconstruction game, a training-free evaluator that decomposes compliance judgments into constrained semantic comparisons. Given only a response, an LLM reconstructs the request answered by its substantive content, excluding information merely repeated or refused. The game compares this reconstruction against similar decoy requests to determine whether it matches the original request, allowing abstention when neither candidate matches. Across three off-the-shelf LLM judges, the game achieves the lowest mean error against human labels among evaluated methods and reduces variation in reported ASR. It prioritizes avoiding false positives, even at the cost of missing some successful jailbreaks.
Read the full abstract ▾
Computer Vision
Ngoc Duy Tran, Kien Do, Duc Thanh Nguyen et al. (+1 more)
One-step
diffusion distillation achieves strong performance on small
datasets but remains insufficient for high-resolution or complex text-to-image settings. Extending to multistep
distillation often requires training data, suffers from early-step degradation, inconsistent outputs across sampling budgets, and incurs high computational cost. We identify that these limitations stem from simultaneous training, where a single adapted network approximates multiple score functions of different student distributions at once, causing biased distribution matching and unstable optimization. Therefore, we propose Progressive Multistep Data-free
Distillation (PMDD), which assigns a dedicated adapted network per student distribution while avoiding the
memory overhead of maintaining multiple adapted networks. Combined with a stage-wise unforget loss to preserve earlier-step generations, PMDD reduces optimization bias in the student, provides faster convergence, and a single unified model that supports multiple inference budgets. Extensive experiments show that PMDD matches or outperforms data-free few-step baselines at every multistep budget, reaches FID 1.99 on ImageNet 64x64 in two steps, below its 512-NFE teacher (2.32), and achieves FID 7.76 on zero-shot COCO 512x512, while requiring 4-5x less training compute than data-free DMDv2 to support three inference budgets.
Read the full abstract ▾
All authors Ngoc Duy Tran, Kien Do, Duc Thanh Nguyen, Truyen Tran
Quantum Physics
Leakage benchmarking with decohering subgroups in the presence of preparation and measurement errors
Shawn Geller, Tongyu Zhao, Alex Kwiatkowski et al. (+3 more)
A common problem with physical
qubits is that the state can leak outside the computational subspace. To characterize the problem, it is desirable to measure both the leakage probability and the fidelity of gates inside the computational subspace. Randomized benchmarking is a readily accessible method for estimating average gate fidelities and can be modified to account for leakage. However, in the presence of state preparation and measurement errors (SPAM) there are parameter degeneracies, and fidelities and leakage become underdetermined. We study the example of qudits with any number of leakage levels under the assumption of gate-independent noise. We show that randomized benchmarking sequences with leakage level decoherence can be analyzed as a Markov process with noisy classical preparations and measurements. This result generalizes to randomized benchmarking with random gates from an arbitrary group that contains a phase-decohering subgroup. We characterize the parameter degeneracy in inferring the Markov process, preparations and measurements. Based on this characterization we develop a measurement and inference method taking advantage of known conservative bounds on SPAM to constrain fidelity, leakage and other parameters of the Markov process, state preparations and measurements. As part of our study, we experimentally implemented the measurement method for a
qubit defined on the one-excitation subspace of two transmons, for which there are two leakage levels. We determined confidence intervals accounting for parameter degeneracy for a number of performance parameters of the system. For leakage and fidelity, we compare these intervals to those obtained for several conventional methods for leakage benchmarking.
Read the full abstract ▾
All authors Shawn Geller, Tongyu Zhao, Alex Kwiatkowski, Scott Glancy, Raymond W. Simmonds, Emanuel Knill
Analysis of PDEs
Relative Entropy and the Inviscid Limit for Planar Shocks of the Multi-D Burgers' Equation
Geng Chen, Jeffrey Cheng, Alexis F. Vasseur et al. (+1 more)
In recent work (arXiv:2610.10341), the authors introduced a framework for studying the vanishing viscosity limit for sequences of weak solutions of the multi-dimensional compressible Navier-Stokes equations to planar shocks of the compressible Euler equations. In this paper, we introduce these methods and apply them to the multi-dimensional viscous Burgers' equation. We show that arbitrarily large viscous shocks enjoy $L^2$-stability among perturbations whose gradients in the directions transverse to the shock are controlled in an appropriate manner. As an application, we study the rate of convergence to the inviscid shock in the vanishing viscosity limit.
Read the full abstract ▾
All authors Geng Chen, Jeffrey Cheng, Alexis F. Vasseur, Yi Wang
Machine Learning
DTDD: Divergence-Triggered Dynamic Distillation for Reliable On-Policy Supervision
Qijia He, Zequn Li, Spencer Luo et al. (+6 more)
On-policy
distillation (OPD) provides dense
token-level supervision on student rollouts, but this supervision can become unreliable when the student and teacher diverge, either because of a large initial policy gap or compounding errors in long-horizon
reasoning and tool use. We introduce Divergence-Triggered Dynamic
Distillation (DTDD), an adaptive OPD framework that uses segment-level student-teacher divergence to trigger teacher takeover during training. When divergence triggers an intervention during
reasoning, the student prefix continues to receive standard OPD supervision. The teacher then generates a recovery suffix, which the student learns through behavior cloning, using cross-
entropy on the teacher-generated
tokens. We evaluate DTDD in three regimes: long-horizon
reasoning,
distillation from a base initialization without offline supervised fine-tuning (SFT), and multi-turn
agents. We also compare alternative trigger designs and teacher-intervention rates. DTDD reduces the post-warmup peak gradient norm by $8.6\times$ and, from a base initialization, reaches $21.3\%$ held-out math accuracy in $2.6\times$ fewer optimizer updates than OPD; both methods finish near $22\%$. It improves best-observed performance by $4.8$ and $4.1$ percentage points on ALFWorld with LoRA and agentic code, respectively. Ablations show that segment-averaged triggers outperform
token-spike triggers. Behavior cloning is also more effective than the tested OPD surrogate for training teacher-generated recoveries. Across 23 experimental arms, takeover tends to help when the teacher substantially outperforms the OPD-trained student, provided that the teacher can recover from student-visited states, the student can learn those recoveries, and teacher intervention is neither excessive nor insufficient.
Read the full abstract ▾
All authors Qijia He, Zequn Li, Spencer Luo, Tianle Chen, Pradyumna Narayana, Zixian Ma, Ranjay Krishna, Lan Nie, Sugato Basu
Artificial Intelligence
Beyond ICD Codes: Semantic IDs for Generative Medical Coding
Yihang Cheng, Qian Zhang, Yonghui Huang et al. (+2 more)
Generative approaches to automatic International Classification of Diseases (ICD) coding use large
language models (LLMs) to predict ICD codes from clinical documents. However, original ICD code strings are not ideal generation targets because the ICD hierarchy does not always reflect clinical relationships, and LLMs struggle to interpret and generate these codes. We introduce SemICD, a framework that trains LLMs to generate Semantic IDs (SIDs) instead. SemICD constructs SIDs by hierarchically clustering
embeddings of textual code profiles, allowing semantically related codes to share prefixes while each complete SID uniquely identifies one ICD code. To align the SID
tokens with clinical semantics, we employ two auxiliary tasks for bidirectional generation between textual code profiles and SIDs. Experiments on MIMIC-III and MIMIC-IV with four LLMs show that SemICD outperforms direct code generation, improving Micro-F1 by 1.34-8.00 points. Our code is available at https://github.com/youwillbethebest/SemICD.
Read the full abstract ▾
All authors Yihang Cheng, Qian Zhang, Yonghui Huang, Baoshuo Kan, Junyi Shen
Information Theory
The Capacity of the Noisy Torn Paper Channel and the Noisy Shuffling Channel
Junsheng Liu and Netanel Raviv
The torn paper channel breaks a codeword into fragments and returns all fragments without their original order. The noisy torn paper channel additionally introduces bit substitution errors to each fragment. Recovering the message therefore requires accounting for both bit errors and uncertainty about the fragment order. We establish a capacity upper bound for the noisy torn paper channel that matches the achievable rate of Ravi, Vahid, and Shomorony and resolves their
conjecture. Our results also determine the capacity of the noisy shuffling channel of Shomorony and Heckel, in which all fragments have the same length, and resolve their open question.
Read the full abstract ▾
Quantum Physics
Gap-Priced $|T\rangle$-States: Dynamic $|T\rangle$ Post-Selection and Budget-Aware Factory Provisioning
Maxwell Poster, Sayam Sethi, Yuhao Liu et al. (+2 more)
Magic $|T\rangle$-state preparation dominates the cost of fault-tolerant
quantum computation on surface codes, yet cultivation and
distillation are typically provisioned statically, leaving much of the program's error budget unused. We present Gap-Priced $|T\rangle$ States (GP$|T\rangle$-S), a
lattice surgery scheduler and factory allocator that co-optimizes both methods under a total $T$-error budget. GP$|T\rangle$-S uses the complementary gap of each cultivated state as a fidelity estimate and dynamically adjusts the cultivation gap cutoff, converting budget slack into speed by giving the scheduler control over the cultivation post-selection. GP$|T\rangle$-S also characterizes
distillation factories by size, $|T\rangle$-state production rate, and budget slack, and introduces cultivating factories that embed cultivation within routing channels, achieving fidelities otherwise requiring two-level
distillation at a single-level footprint. GP$|T\rangle$-S organizes these characterizations into a factory catalog, and performs
distillation factory allocation over this catalog in tandem with the gap cutoff using dynamic-programming to achieve one of i) minimum spacetime volume, ii) minimum execution time, or iii) minimum space overhead, all within a user-provided $T$-error budget. Across sixteen
benchmarks under high and low $T$-error budgets of $10^{-3}$ and $10^{-5}$ respectively, GP$|T\rangle$-S consistently outperforms cultivation-only and single-factory baselines without exceeding the budget, achieving a 5.5$\times$
geometric mean speedup in logical cycle count and a 3.4$\times$
geometric mean reduction in space-time volume.
Read the full abstract ▾
All authors Maxwell Poster, Sayam Sethi, Yuhao Liu, Jonathan Baker, Costin Iancu
Machine Learning
Verifier-Grounded Agentic RL for Provably Correct Code
George Ma, Yuhao Zhang, Bing He et al. (+5 more)
Reinforcement learning for code usually rewards passing tests, which a model can do while remaining wrong on the inputs the tests do not cover. A program verifier instead decides whether a program satisfies a formal specification for all inputs.
Language models are almost always paired with verifiers at inference time only, and turning a verifier into a reward raises two difficulties. It reports only acceptance or rejection, which gives no gradient when every attempt fails, and a policy can cheat it, most simply by telling it to assume what the proof was supposed to establish. We build an agentic environment and reward that address both. The reward reads the verifier's diagnostics rather than its verdict, scoring a candidate by the weighted fraction of proof obligations it discharges, extended with hidden probes from a reference proof's annotations. An integrity audit outranks the verifier, rejecting any candidate that alters the specification or assumes what it was meant to prove. The model works as an
agent against a live verifier over many turns, with a separately trained proof specialist as a frozen collaborator. Across Qwen3 models from 1.7B to 14B and a 35B Qwen3.6 mixture-of-experts model, held-out verification on code generation rises from 66.1% to 93.5% on DafnyBench and from 56.3% to 100% on SWE-Proof. This reward leads every other training signal we compare, including a binary pass-or-fail reward trained under otherwise identical settings, with the widest margin on code generation. When the
agent must also deliver the repository patch, graded by each project's withheld tests, the fraction of held-out issues it resolves rises from 59.6% to 72.5%. Evaluated unchanged on three verification suites withheld from training, the checkpoints improve on all of them, including one whose
language and verifier are absent from training.
Read the full abstract ▾
All authors George Ma, Yuhao Zhang, Bing He, Zeren Shui, Ferhat Erata, Murali Krishna Ramanathan, Somayeh Sojoudi, Hao Zhou
Solar and Stellar Astrophysics
Twisted Pseudodisk and Asymmetric Mass Accretion on the Circumstellar Disk II: Parameter Survey
Masahiro N. Machida, Shingo Hirano and Shantanu Basu
Using three-dimensional magnetohydrodynamic simulations accounting for ohmic dissipation, we investigate the long-term
evolution of the infalling envelope for $\sim 100,000$\,yr after protostar formation. The primary control parameter is the initial misalignment angle $\theta_0$ between the rotation axis and the
magnetic field, varied from $0^\circ$ to $90^\circ$. In all misaligned models ($\theta_0 \ne 0^\circ$), the pseudodisk (or infalling envelope) surrounding the circumstellar disk develops a highly distorted and complex morphology. Gas
accretion from the envelope proceeds through discrete, channel-like flows, producing asymmetric streams rather than axisymmetric infall. Protostellar outflows are launched even in these complex environments, and inflow and outflow regions are strongly interwoven. Outflows transport angular momentum anisotropically, leading to time-dependent changes in disk orientation and outflow direction, which further enhance envelope asymmetry. We examine the relative orientation between gas flows and
magnetic fields and find that, although statistical tendencies exist,
accretion streams cannot be uniquely characterized as either parallel or perpendicular to the
magnetic field, particularly in strongly misaligned cases. We also show that most infalling gas accretes at nearly the free-fall velocity, with only a minor fraction significantly slowed within the pseudodisk. Asymmetric
accretion and stream-like structures persist beyond 10,000\,au, demonstrating that large-scale asymmetry can arise intrinsically. Our results show that complex streams and asymmetric envelopes naturally emerge from magnetized core physics, even in the absence of initial
turbulence or external gas
accretion.
Read the full abstract ▾
Numerical Analysis
Variationally correct loss functions for nonlinear parametric problems and their neural approximations
Pablo Cortés Castillo, Wolfgang Dahmen and Jay Gopalakrishnan
We construct computable a posteriori error
estimators for parametric families of nonlinear elliptic problems by measuring the residual of a variational formulation in the dual norm of a closed subspace of the test space. A single abstract
theorem shows that the resulting
estimator is bounded above and below by the error in the trial norm, i.e., it is a variationally correct loss function. Two instances are covered. The first considers an $L_2$-like test space, yielding a nonlinear first-order system least-squares loss function that certifies arbitrary, possibly meshless, approximations. The second breaks the test space, yielding a nonlinear discontinuous Petrov-Galerkin loss function computed by element-local problems. All assumptions are verified for quasilinear
diffusion, including monotone fluxes and semilinear problems, and a nonlocal Kirchhoff model. Numerical studies train neural networks approximating the parameter-to-solution maps with the latter loss and measure the ratio of loss to squared error.
Read the full abstract ▾
Robotics
Learning Keypoint Constraints from Visual Demonstrations for Safe Motion Planning
Zheng Qiu, Zhouyu Zhang, Ninghan Zhong et al. (+1 more)
We present Keypoint Constraint Learning (KCL), a framework for learning task-relevant motion constraints from visual demonstrations and for planning safely with the extracted constraints. To make constraint learning tractable from RGB-D observations, we represent demonstrations using sparse 3D object-centric keypoints and semantic mask centroids extracted through segmentation, visual feature clustering, and point tracking. In this reduced
geometric space, we learn keypoint dynamics and infer distance- and orientation-based constraints that make the expert trajectories feasible and approximately locally optimal by enforcing KKT optimality conditions in the reduced state space. We then plan with the learned constraints under uncertainty, using robustness margins derived from accumulated prediction error, and provide a theoretical analysis of how keypoint perception error and learned dynamics error affect the recovered constraints. We evaluate KCL on visual demonstrations from navigation to rigid/deformable manipulation, showing that it recovers multi-stage
geometric constraints and outperforms baselines in downstream safe planning.
Read the full abstract ▾
All authors Zheng Qiu, Zhouyu Zhang, Ninghan Zhong, Glen Chou