Computer Vision
Targeted Visual Counterfactual Explanations for Contrastive Vision-Language Model
Van Bach Nguyen, Jörg Schlötterer and Christin Seifer
Current explanation methods for contrastive
vision-
language models such as CLIP mainly identify important regions without showing how to change the input in order to get a target prediction. We introduce Mask-guided Adaptive Counterfactual Explanations (MACE), a targeted visual counterfactual method designed specifically for CLIP zero-shot classification. MACE constructs an editable region from either source attribution or source-target attribution differences and expands the mask only when needed to reach a specified target class. A latent
diffusion inpainting model then modifies the selected region, while a frozen CLIP model provides modification guidance and anchors the remaining image content to the original input. We evaluate MACE on ImageNet, Food-101, Oxford Pets, and CUB-200. The source-mask variant achieves the highest target top-1 success rate across all four
datasets, while the difference-mask variant produces the smallest pixel level and perceptual changes and the best realism scores. Both variants improve proximity and realism over a Stable
Diffusion-only baseline using the same generative backbone. These results show that adaptive mask-guided editing produces effective CLIP counterfactuals. They further reveal a tradeoff between counterfactual validity and source-image preservation.
Read the full abstract ▾
Physics and Society
Statistical Physics of Fish Collective Motion
Elena G. de Lamo, M. C. Miguel and R. Pastor-Satorras
We review recent theoretical and empirical advances that recast fish schooling within a statistical-physics framework, emphasizing how heterogeneous and time-dependent interactions govern collective motion. Building on extensions of Vicsek-type models to complex and weighted social networks, we show that
topology and link strength qualitatively alter flocking stability and critical thresholds, with empirical weights typically reducing global
alignment. High-resolution trajectory inference reveals selective, nonreciprocal responses: individuals preferentially attend to faster neighbors, producing transient, speed-induced leadership rather than fixed hierarchies. At the mesoscopic scale, schools exhibit avalanche-like turning cascades with scale-free size and duration statistics, aftershock clustering, and an Omori-type temporal decay with a short
memory. Together, these findings support a unified picture in which collective order and critical-like fluctuations coexist:
alignment provides stability while near-critical variability preserves responsiveness. This synthesis highlights fish schools as a tractable experimental model for how living collectives organize, transmit and process information across scales.
Read the full abstract ▾
Quantum Physics
Resonance Breaking in Noisy Shor's Algorithm
Zhengwei Liu
We argue that the exponential
quantum advantage in Shor's
algorithm is broken under one-layer depolarizing noise of arbitrary small error rate, by analyzing a resonance breaking phenomenon in noisy
quantum circuits. First, we express the distribution of measurements on $n$-
qubit strings as the superposition of the $4^n$ wave functions in the Pauli path integral. In the noiseless case, the bit-strings achieving resonant peaks guaranteed a constant rate of successful measurements to factor a large number. Secondly, when the middle layer of the
quantum circuit has independent depolarizing noise of rate $\lambda$, for any Hamming weight $d$, we obtain corresponding measurement rate bounded by $2(1-\lambda)^d$ for higher frequency terms and by $O(n^d)/2^{n/2}$ for low frequency terms. The rate approaches to zero as $n$ and $d$ approaches to infinity. Resonance breaking destroys the exponential
quantum advantage. Furthermore, we design a classical factoring
algorithm in
polynomial time $O(n^d)$ to substitute the low frequency contribution.
Read the full abstract ▾
Computation and Language
Co-Linguistics: AI-augmented Theory Construction in Linguistics
Emmanuel Chemla, Benjamin Spector, Alexandros Kalomoiros et al. (+1 more)
LLMs have been studied in recent linguistics as potential models of humans' linguistic abilities. Here we discuss an entirely different use of AI, namely as a co-scientist, to help construct and assess linguistic theories (we refer to the result as "Co-Linguistics"). Since the 1960s, linguistics has developed theories that are in principle mathematically formalizable, often in the
language of formal
language theory or model theory. The AI revolution in mathematics will thus have consequences in linguistics-but with an essential twist: proving new
theorems is rarely the linguist's goal. Rather, one seeks to find the best set of axioms to derive empirical statements. AI could accelerate research by making existing theories fully explicit, by comparing competing theories, and more ambitiously, by proposing new theories (in machine learning, this relates to "program induction"). It will also help assess theories by accelerating the identification and test of crucial predictions, thanks to unparalleled access to data (in machine learning, this relates to "active learning"). While the cycle from theory evaluation to theory construction may give rise to recursive and possibly autonomous improvement of linguistic theories, humans remain central: linguists provide scientific directions and evaluate theories conceptually, and experimental participants are needed to assess empirical predictions that are outside the reach of LLMs.
Read the full abstract ▾
All authors Emmanuel Chemla, Benjamin Spector, Alexandros Kalomoiros, Philippe Schlenker
Algebraic Geometry
On the Hilbert polynomial of the linked projective space
Felipe De León, Eduardo Esteves and Eduardo Vital
Linked projective spaces are quiver Grassmannians of subspaces of dimension 1 of certain quiver representations. Degenerations of linear series produce these representations, with the limit divisors parameterized by the associated linked projective spaces. It is not known whether all linked projective spaces arise this way. If they do, they are degenerations of the (small) diagonal in a product of projective spaces. In any case, we prove here that they have the (multivariate)
Hilbert polynomial of the diagonal. To achieve this, we first extend a
Hilbert-
polynomial formula for multiplicity-free varieties to (simple) normal-crossings schemes with multiplicity-free strata in products of projective spaces. Then we prove that a linked projective space is normal-crossings, by describing it locally in terms of Mustafin varieties. Finally, we use a relation between intersections of components of the linked projective space and certain polytopes in the tiling of a simplex associated to the linked net to prove that our formula for the
Hilbert polynomial applies.
Read the full abstract ▾
Machine Learning
RLTLDR: Self-improvement by Internalizing Self-generated Feedback
Michael Kirchhof, Eleonora Gualdoni, Andrew Szot et al. (+6 more)
The common paradigm of reinforcement learning with verifiable rewards (RLVR) is to let
agents make multiple attempts at a task, and optimize towards the successful ones. This becomes problematic in the realms of self-improvement, where tasks are so difficult that the
agent has a low or even no chance of success, and where there are no teacher models or example solutions to distill from. In this paper, we introduce RLTLDR. After each failed attempt, we show the policy the verifier outputs and let it write its own feedback, in the form of a single TL;DR insight. The next rollout is conditioned on all previous insights, and we sequentially sample rollouts until a solution is found. Moreover, we enable backpropagation on the in-context insights to internalize a direct task to insight mapping. On challenging tool-calling and coding
datasets (filtered to Pass@128=0), standard GRPO training of a Qwen 3.5 9B Thinking policy stays flat at a Pass@1 of 0% to 1%. RLTLDR breaks through this learning barrier, achieving a Pass@1 of 14-31% with insights in context during training and, crucially, 12-13% when no insight is in context at eval time. We identify that the key is the task to insight internalization. To study this further, we reduce our approach to SFTLDR, training only on (task, insight) tuples, without showing or backpropagating on any rollouts. Training on only 4k of these tuples recovers almost the full performance of RLTLDR and classical SFT on full rollouts. This demonstrates a promising compacted training paradigm of the form "on this sort of task, keep this sort of thing in mind", which we hope to inspire future research on.
Read the full abstract ▾
All authors Michael Kirchhof, Eleonora Gualdoni, Andrew Szot, Khashayar Gatmiry, Aryo Lotfi, Abbas Kazerouni, Omar Attia, Sanjoy Chowdhury, Alexander Toshev
Machine Learning
ProCTI: Prototype-Refined Global Conditioning for Diffusion-Based Time Series Imputation
Fariza Rashid, Duc Van Le, Rahat Masood et al. (+3 more)
Time series imputation has progressed from statistical and deep learning approaches to
diffusion-based models, which have shown strong recent performance. Existing
diffusion-based methods typically condition the reverse process using local contextual information from the current or neighbouring windows. Meanwhile, global
dataset-level structure often remains implicit, limiting performance when local observations are sparse, noisy, or unrepresentative. To address this issue, we propose ProCTI, a
diffusion-imputation framework that augments local conditioning with retrieved global
dataset-level priors through learned prototypes. A hybrid conditioning mechanism integrates this global context with local signals during reverse
diffusion, enabling more accurate reconstruction under varying missingness scenarios. Experiments across multiple
benchmark datasets show that ProCTI outperforms strong baselines overall under random missingness, while remaining competitive under attribute-wise missingness. Furthermore, we use a latent-regime data model to characterise the precise conditions under which prototype-derived global conditioning provably improves imputation. We support this with a general theoretical analysis of local-global conditioning.
Read the full abstract ▾
All authors Fariza Rashid, Duc Van Le, Rahat Masood, Gustavo Batista, Aruna Seneviratne, Suranga Seneviratne
Machine Learning
Procedural Core: A Compact Recurrent Initialization for Vision Transformers
Zachary Shinnick, Christian Internò, Hemanth Saratchandran et al. (+2 more)
Transformers are typically trained from random initialization, requiring all their capabilities to emerge from large-scale optimization. Recent work showed that a small amount of abstract procedurally generated data can help acquire generic inductive structure at low cost. However, this adds a pretraining stage that must be repeated for every target model. We propose Procedural Core, an initialization strategy that captures this generic structure into a compact set of weights that can be reused across models. We train a minimal recurrent
transformer on procedural data, then expand its weights to initialize
transformers of arbitrary width and depth. The resulting initialization improves performance on image classification, self-supervised visual learning (DINO), and modeling natural
language (FineWeb-Edu) and code (CodeParrot). For image classification, expanding a 1M-parameter core to initialize an 85M-parameter ViT-Base improves ImageNet top-1 accuracy by 2.2 pp over standard random initialization. Our analysis identifies recurrence as essential for learning compact weights that transfer across models. In ViTs, we localize a key benefit in the suppression of high-norm
tokens that produces substantial improvements in zero-shot
segmentation (ImageNet-S mAP 32.3 to 42.9), object localization (VOC07 CorLoc 9.9 to 18.4), and depth estimation (NYUv2 RMSE 1.104 to 0.998). This demonstrates that
transformers need not start from a blank slate, and can be initialized with generic capabilities at low cost with no domain- or task-specific data.
Read the full abstract ▾
All authors Zachary Shinnick, Christian Internò, Hemanth Saratchandran, Anton van den Hengel, Damien Teney
Statistical Methodology
On the use of G-computation in small randomized controlled trials with many covariates
Muluneh Alene, Kelly Van Lancker and Stijn Vansteelandt
In contemporary randomized controlled trials (RCTs), the number of patients is often small relative to the number of baseline
covariates collected. In such settings, maximum likelihood
estimators of conditional treatment effects in generalized linear models, along with their standard errors, may exhibit substantial bias. This study examines whether similar bias arises in G-computation
estimators of marginal treatment effects, which are known for their robustness to model misspecification when applied with canonical GLMs. We develop theoretical insights, drawing on recent literature on G-computation under proportional asymptotic regimes, in which the number of
covariates grows with the sample size. Specifically, we characterize the bias of the standard G-computation
estimator and the leave-one-out G-computation
estimator under such high-dimensional settings. Monte Carlo simulations using linear and logistic outcome models are conducted to evaluate practical remedies for G-computation
estimators, including
covariate selection, cross-fitting, leave-one-out cross-fitting, and small sample corrections to standard errors. Further insights are derived from a re-analysis of the BestAIR trial data. Our findings provide guidance for the application of G-computation in modern RCTs, particularly when addressing challenges posed by limited sample size. Keywords:
Covariate adjustment,
covariate selection, high-dimensional, proportional asymptotics, targeted learning, cross-fitting.
Read the full abstract ▾
Quantum Physics
Adaptive decoding of quantum LDPC codes through decoder disagreement
Maida Wang and Peter V. Coveney
Accurate decoding of
quantum low-density parity-check (qLDPC) codes often relies on expensive post-processing search, although decoding difficulty varies strongly between syndromes. We find that the benefit of deeper post-processing search is highly concentrated in a small subset of decoding instances, and that these instances can be identified directly from the decoder itself. To this end, we introduce an adaptive decoder based on belief propagation (BP) and ordered-statistics decoding (OSD), using the disagreement between the BP hard decision and the syndrome-consistent order-zero OSD solution as an internal risk signal to determine where deeper search is required. On a $[[144,12,12]]$ bivariate bicycle code under
circuit-level depolarising noise, encoding $12$ logical
qubits in $144$ data
qubits with distance $12$, escalating only the highest-risk $20\%$ of instances recovers $85\%$ to $92\%$ of the improvement in logical error rate available from a full one-free-variable sweep, while reducing the mean serial decoding cost by a factor of $3.6$ relative to applying the same truncated search to every instance. At matched budget, disagreement-guided routing also outperforms routing based on the BP syndrome residual, syndrome weight and random selection. The same behaviour reappears on a structurally distinct radial qLDPC code obtained from a lifted-product construction, where escalating $20\%$ of instances recovers $87\%$ to $94\%$ of the available gain. A $Z$-
memory experiment on the Quantinuum H2 trapped-ion processor, implementing both $X$-type and $Z$-type checks, further shows that the signal remains predictive under device noise. These results show that decoder-internal disagreement can expose where additional decoding effort is valuable, allowing classical computation to be concentrated on the instances most likely to benefit from it.
Read the full abstract ▾
Quantum Physics
Operational theory for photonic circuits: generalizing linear optics beyond quantum theory
Ismaël Septembre, Matthias Kleinmann and Martin Plávala
We use the analogy between the quadrature operators in
quantum optics and the phase-space coordinates in a mechanical harmonic oscillator to generalise the theory of linear optics beyond
quantum theory. For this, we introduce a framework for analysing photonic
circuits within generalised probabilistic theories based on quadrature operators. Using this framework, we ask whether phenomena like the Hong-Ou-Mandel effect, Mach-Zehnder interferometry are specific to
quantum theory or also occur across a broader class of theories. We find the latter: the vanishing
photon coincidences at the output of a beam splitter also occur for classical light and for a modified version of quantised light. Similarly, for Mach-Zehnder interferometry we find that there is no qualitative difference between
quantum theory and these alternative theories. Our framework can serve as a foundation to study generalised theories of complex photonic set-ups and photonic
quantum computers.
Read the full abstract ▾
Networking
RingStitch: Demand-Aware Optical Stitching for Fragmented TPU Clusters
Huiru Ao, Fan Yang, Binglei Wang et al. (+2 more)
Large-scale AI training clusters increasingly use optical
circuit switching (OCS) to reconfigure rack-level interconnects and create elastic accelerator slices. In multi-tenant TPU-style clusters, however, small and medium jobs often leave partial free capacity stranded inside racks. Although the aggregate free capacity may be sufficient for a new job, it cannot be used by local placement or coarse full-rack stitching. This paper presents RingStitch, an OCS-based defragmentation scheduler that turns fragmented rack capacity into schedulable resources. RingStitch follows a local-first policy, stitches compact cross-rack fragments only when needed, and orders the selected racks into a low-cost logical ring. Simulations on a TPU 8t-like SuperPod model show that RingStitch improves schedulability over local and full-rack baselines.
Read the full abstract ▾
All authors Huiru Ao, Fan Yang, Binglei Wang, Bo Liu, Jialong Li
Distributed Computing
SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving
Chuan Liu, Shuoming Zhang, Zhicheng Li et al. (+6 more)
No single way of parallelizing
attention serves large
language models well under all loads. Low concurrency favors tensor parallelism, many independent requests favor data-parallel
attention, and long
prompts favor context parallelism.
Reasoning, agentic, and RL-rollout workloads make a fixed choice untenable: a batch that begins as many short requests ends as a few very long ones, so the best layout changes while the same requests run. Serving engines nevertheless fix one layout at launch, because changing it has meant draining requests and restarting workers. We present SPLASH, a serving system that switches the parallel layout of
attention while requests are running. It builds on one observation: modern
attention, with few or no KV heads, decouples where a request's KV cache lives from how
attention weights are sharded. This has two consequences. First, layouts differ only in who owns the weights and the cache, and most of that state already sits where the next layout needs it; SPLASH reuses it, moves the rest in the background of ongoing inference, and hands off at a batch boundary, making a switch nearly free: its median overhead is under 0.51% of the step it runs in. Second, the decoupling exposes a layout that existing engines lack: Decoupled Ownership Parallelism (DOP) shards
attention weights as tensor parallelism does while keeping each request's cache on a single owner as data-parallel
attention does. DOP replicates neither, offers 27-60% more KV capacity than data-parallel
attention, and gives the scheduler a choice when KV
memory limits admission. A transition-aware scheduler follows the best of the four layouts as load changes. On B200 GPUs serving GLM-5.3, SPLASH improves end-to-end serving throughput by 1.3-1.73x over fixed-layout deployments, and the same layout regimes appear with DeepSeek-V3.2 on H200 and GLM-5.3-Flash on DCU.
Read the full abstract ▾
All authors Chuan Liu, Shuoming Zhang, Zhicheng Li, Qianqi Sun, Ruiyuan Xu, Qiuchu Yu, Xiyu Shi, Huimin Cui, Jiacheng Zhao
Astrophysics of Galaxies
GA-NIFS: sRMS, a new method for disentangling resolved and unresolved emission in integral field spectroscopy. Application to distant quasars
Michele Perna, Santiago Arribas, Carlota Prieto-Jiménez et al. (+28 more)
The interpretation of AGN emission-line profiles in type1 AGN is often complicated by the dominance of bright, spatially unresolved
nuclear continuum and broad-line region (BLR) emission over narrower and blended components from the narrow-line region (NLR) and the host
galaxy. This remains challenging even with integral-field spectroscopy (IFS), as the relevant spatial variations can occur on scales smaller than the angular resolution, particularly in the compact host of distant AGN. We introduce the spatial root-mean-square (sRMS) technique, a new approach to analyse single-epoch IFS that uses the spatial variance of spectra extracted from partially overlapping apertures to isolate off-
nuclear emission of high-z BL-AGN, and investigate whether the resulting kinematic information can improve the decomposition of their integrated spectra. We construct sRMS spectra from ensembles of overlapping apertures centred on the unresolved
nucleus. Emission that is spatially invariant on the scales probed by the apertures is consequently suppressed, whereas spatially varying emission is retained. We apply the procedure to JWST/NIRSpec data of two z~6.5 QSOs. We model the resulting sRMS spectra with multiple Gaussian components to determine the kinematics of the spatially varying emission, and use these kinematic constraints to model the integrated spectra. The sRMS technique identifies spatially varying narrow-line emission on scales up to ~5x smaller than the NIRSpec PSF and provides robust constraints on the kinematics of individual kinematic components. Imposing these kinematic constraints on the integrated-spectrum fit reduces significantly the degeneracy of multi-component decompositions. Simulations further demonstrate the potential of the method to detect spatially offset BLR emission from close dual BL-AGN, down to projected separations of ~200 pc at z~6.5 for a BLR flux ratio of ~10.
Read the full abstract ▾
All authors Michele Perna, Santiago Arribas, Carlota Prieto-Jiménez, Isabella Lamperti, Lorenzo Ulivi, Hannah Übler, Giovanni Cresci, Bruno Rodríguez Del Pino, Torsten Böker, Andrew J. Bunker, Stéphane Charlot, Roberto Maiolino, Chris J. Willott, Ivan Delvecchio, Elena Bertola, Madeline A. Marshall, Eleonora Parlanti, Pablo G. Pérez-González, Giacomo Venturi, Sandra Zamora, Filippo Mannucci, Quirino D'Amato, Alessandra De Rosa, Ismael García-Bernete, Alessandro Marconi, Miguel Pereira-Santaella, Fabio Rigamonti, Martina Scialpi, Paola Severgnini, Cristian Vignali, Maria Vittoria Zanchettin
Computation and Language
Correct, Don't Delete: Mitigating Emergent Misalignment with Corrective Supervision
Jacob Epifano
Fine-tuning a
language model on a narrow set of harmful demonstrations, such as bad medical advice, can make it broadly misaligned on unrelated questions, a phenomenon known as emergent misalignment (EM). The usual defense is to find the offending rows and delete them, but a row locator failed our held-out test and deleting rows helps less than expected. We ask a different question: given a fixed set of poisoned rows, is it better to correct them than to remove them? We fine-tune Qwen2.5-14B-Instruct on a mixture of bad medical advice and benign chat data, select a quarter of the poison rows in advance, and either delete them or replace each with a corrected answer to the same
prompt, keeping everything else the same. Replacing the rows cuts the EM rate by about a third and improves answers on held-out medical questions, while deleting the same rows has little measurable effect. The advantage is larger when half the poison rows are corrected, and it holds on a second base model and a second misaligned model organism. The content of the replacement appears to matter: paraphrasing the rows while keeping their bad advice shows no clear benefit, and the correct answers distributed with the
dataset appear to do about as well as our rewriter's. Realigning an already-poisoned model with further fine-tuning is known to work, but which data does the work has not been compared directly. We find that a short round of training on corrections beats the same amount of training on generic chat data, that corrections on other medical
prompts do roughly as well as corrections of the poisoned
prompts themselves, and that instructing the correction writer to model a careful, harm-avoiding assistant adds no measurable benefit over plain corrections. In the settings we tested, correcting harmful training data reduces EM more than deleting it.
Read the full abstract ▾
Medical Physics
Using a radiopharmaceutical production cyclotron for radiobiological research
Eva Kasanda, Jan Gruber, Pierluigi Casolaro et al. (+8 more)
The Laboratory for High Energy Physics and the Institute of Anatomy at the University of Bern are establishing a pre-clinical
proton therapy research facility at the Bern University Hospital (Inselspital). The facility centers on the Bern Medical Cyclotron, an 18 MeV compact cyclotron designed for microampere-range beam currents for radioisotope production. The Bern Medical Cyclotron can be adapted as a highly accessible and customizable platform for pre-clinical
proton therapy research, enabling advanced modalities such as ultra-high dose rate (FLASH) radiotherapy and spatially fractionated radiation therapy. Human keratinocytes (HaCaT) and mouse melanoma (B16-F10)
cell lines were irradiated with 8.14(29) MeV
protons at conventional dose rates and with 225 kV X-rays at doses from 0 to 8 Gy. Clonogenic survival assays were analyzed using the linear-quadratic model, and relative biological effectiveness values were derived by comparing the
proton and X-ray doses required for 0.1 survival.
Proton irradiation yielded steeper survival curves than X-rays, indicating higher cytotoxicity. B16-F10
cells were most sensitive, with a relative biological effectiveness of 1.34 at 0.1 survival, while HaCaT
cells showed a relative biological effectiveness of 1.21. These results confirm enhanced
proton effectiveness in
cell killing and demonstrate that the Bern Medical Cyclotron provides a reliable platform for radiobiological studies. Establishing in vitro
proton irradiation capability marks a key milestone toward developing a comprehensive pre-clinical radiobiology research facility in Bern. Ongoing upgrades will further refine dosimetry and
proton delivery, paving the way for future FLASH and
proton spatially fractionated radiation therapy studies.
Read the full abstract ▾
All authors Eva Kasanda, Jan Gruber, Pierluigi Casolaro, Lars Eggimann, Alexander Gottstein, Isidre Mateu, Paolo Pellicioli, Maria Vittoria Rossi, Paola Scampoli, Cristian Fernandez-Palomo, Saverio Braccini
Signal Processing
Joint Localization and Data Detection in Ambient Backscatter Communications: Uncertainty-Preserving Inference and Cross-Frame Geometry Consensus
Xianhua Yu, Dong Li, Bowen Gu et al. (+2 more)
We address joint continuous localization and data detection in ambient backscatter communication, where weak observations can leave competing geometry hypotheses plausible and a staged point-estimate interface can discard information about this ambiguity. We develop the Geometry-aware Frame Network (GeoFrameNet), which projects channel frequency responses onto a physics-derived delay–angle-of-arrival
lattice and aggregates sign-robust evidence across orthogonal frequency-division multiplexing symbols to form a shared geometry posterior over physically feasible candidates. Localization uses the full posterior with candidate-specific refinement; detection combines candidate-conditioned differential logits using posterior weights renormalized over selected candidates. Bit supervision guides geometry scoring. For a fixed device, the Cross-Frame Geometry Consensus Network (CFGC-Net) fuses frozen GeoFrameNet candidate logits and features across frames for localization while preserving frame-wise detection outputs. On an independent test set, GeoFrameNet achieves lower localization root-mean-square error (RMSE) than an all-symbol multiple-measurement-vector sparse
Bayesian learning (MMV-SBL) baseline at all eight evaluated signal-to-noise ratio (SNR) points, with simultaneous bit error rate and RMSE reductions from $-30$ to $-22.5$~dB. At $-30$~dB SNR, the respective RMSEs are 3.1573 and 18.8859~m. CFGC-Net achieves the lowest aggregate RMSE among the evaluated fusion strategies for two, four, and eight frames.
Read the full abstract ▾
All authors Xianhua Yu, Dong Li, Bowen Gu, Tianhao Liang, Tingting Zhang
Functional Analysis
The Stability and Density of Norm Attaining Operators on Reducing Subspaces
Zinan Su and Yuanhang Zhang
For a bounded linear operator on an infinite-dimensional
Hilbert space, we consider the class $\beta(H)$ of operators whose restrictions to all nonzero reducing subspaces attain their norms. We prove that $\beta(H)$ is dense in $\mathcal{B}(H)$ in the operator norm. Since $\beta(H)$ is not stable under arbitrary compact perturbations, we introduce a natural subclass $\beta_0(H)$ and investigate its perturbation properties. We show that $\beta_0(H)$ is stable under finite-rank perturbations and obtain a complete characterization of those compact perturbations that preserve membership in $\beta_0(H)$. We also determine the maximal compactly perturbation-invariant subset of $\beta_0(H)$. These results clarify the spectral structure and stability of norm-attaining operators on reducing subspaces.
Read the full abstract ▾
Quantum Physics
Optimal interferometric certification of multi-photon indistinguishability
Marco Robbio, Michał Oszmaniec, Nicolas J. Cerf et al. (+2 more)
Multiphoton indistinguishability is a key resource for photonic
quantum technologies, yet its characterization typically relies on resource-intensive methods. In this work, we develop two efficient and experimentally friendly protocols to estimate or bound the fidelity $F_{\mathrm{ind}}$ of an $N$-
photon state to the closest perfectly indistinguishable state. The first protocol applies to sources preparing separable states, uses a single
Fourier interferometer together with
photon-number-resolving detection, and yields tight two-sided bounds on $F_{\mathrm{ind}}$. The second protocol combines randomized implementations of linear-optical interferometers with
photon counting, enabling direct estimation of $F_{\mathrm{ind}}$ for arbitrary $N$-
photon states. Both protocols can certify $F_{\mathrm{ind}} = 1 - \mathcal{O}(\epsilon)$ using provably optimal $\mathcal{O}(1/\epsilon)$ samples, in contrast to previous approaches which required prior assumptions on the model of partial distinguishability. Our methods, based on a multiphoton generalization of the Hong-Ou-Mandel test, bring the rigorous and operationally meaningful certification of multiphoton indistinguishability within reach of current photonic technologies.
Read the full abstract ▾
All authors Marco Robbio, Michał Oszmaniec, Nicolas J. Cerf, Ernesto Galvão, Leonardo Novo
Optics
Optical enhancement cavities operated at GHz repetition rate in burst mode regime
Kevin Dupraz, Frédéric Blanc, Ronic Chiche et al. (+7 more)
The design and experimental validation of a burst-mode optical cavity tailored for inverse Compton scattering X-ray sources is presented. The system is tailored for a room temperature linear accelerator operating at 880 MHz and utilizes a common-path injection scheme for both locking and burst
laser pulses to optimize stability and simplicity. We detail the cavity construction, the
laser system, and performance under two enhancement cavity finesse regimes (2000 and 7300), highlighting the sensitivity of cavity filling to carrier-envelope offset parameters. A theoretical model is developed to explain observed power dynamics and to guide optimization for high-energy storage. Our findings demonstrate reliable storage exceeding 300 mJ over 20 minutes, paving the way toward compact, high-flux inverse Compton scattering X-ray sources.
Read the full abstract ▾
All authors Kevin Dupraz, Frédéric Blanc, Ronic Chiche, Aurélien Martens, Antoine Courjaud, Loïc Amoudry, Titouan Le Barillec, Daniele Nutarelli, Yann Peinaud, Fabian Zomer
Numerical Analysis
An augmented mixed finite element method for the Biot problem with nonlinear permeability
Felipe Lepe, Gonzalo Rivera, Ricardo Ruiz-Baier et al. (+1 more)
In two and three dimensions, we analyze a mixed formulation of the nonlinear Biot equations in terms of the poroelastic stress tensor, the displacement, the pressure, and the rotation tensor. To establish the well-posedness of the problem, we introduce suitable stabilization terms, leading to an augmented mixed formulation in which the associated parameters are chosen appropriately. Owing to the nonlinearity of the problem, we employ a fixed-point strategy to prove the existence and uniqueness of solutions. We propose a finite element scheme for which we establish existence and uniqueness of a discrete solution, together with a priori error estimates. Additionally, we design and analyze an a posteriori error
estimator of the residual type which results to be reliable and efficient. Finally, we present a series of numerical experiments in two and three dimensions in order to assess the performance of the proposed method.
Read the full abstract ▾
All authors Felipe Lepe, Gonzalo Rivera, Ricardo Ruiz-Baier, Segundo Villa-Fuentes
Sound
AS$^2$D: Accelerating On-Demand Audio Understanding on Mobile Devices
Yunzhe Li, Kyoungjun Park, Hongzi Zhu et al. (+1 more)
Speculative decoding accelerates autoregressive generation by using a smaller drafter to propose
tokens for batched verification by a larger target. However, conventional speculative decoding couples drafting to the target's evolving verified prefix, serializing drafting and verification. We ask whether this dependency is necessary for source-conditioned generation. Our key observation is that, for audio
language models, the input audio and user request can provide useful speculative candidates without following the target's evolving text prefix. We propose AS$^2$D (Audio Speculative Speculative Decoding), which enables target-decoupled drafting: an audio-conditioned drafter follows its own generation history while the target independently verifies and corrects ready candidates. Without usable candidates, the target advances alone. Thus, target feedback determines which candidates are committed but no longer determines when the drafter can make progress, enabling drafting and verification to proceed concurrently while retaining target-side verification and correction. We implement AS$^2$D in MNN for Android and evaluate two target models across four phones, seven
datasets, and three tasks covering 12.2 hours of audio. Across four phones, AS$^2$D improves pooled ASR throughput by 42-76% over target-only decoding, while only 5.7% of evaluation windows are slower than target-only, compared with 58.1-63.0% for speculative baselines. For ASR, AS$^2$D reaches 97.33-98.20% of a hindsight per-window oracle's pooled throughput over the evaluated drafter/budget catalog. Native on-demand execution with a 7B target achieves up to 78% higher throughput than target-only. These results show that source-conditioned audio generation can relax the conventional dependence of speculative drafting on the target's evolving output prefix, exposing substantial parallelism for efficient inference.
Read the full abstract ▾
All authors Yunzhe Li, Kyoungjun Park, Hongzi Zhu, Lili Qiu
Machine Learning
Authority Bias in Language Models: Source Deference and User Agreement Are Not Interchangeable
Abhinav Rajeev Kumar and Paras Chopra
Language models tend to agree with whatever a user asserts, and post-training increasingly targets this sycophancy so that models evaluate claims on their merits rather than deferring to the user. Yet the same models are far more compliant when a wrong answer is attributed to a verified source, which is how
retrieval results, tool outputs, and grounded-search content often present information. We measure this gap across five open-weight families and three closed APIs. A single verified-source note endorsing a wrong answer flips 45-88% of baseline-correct responses in seven of eight models, and compliance rises with how authoritative the note sounds. Source deference and user agreement are not behaviorally interchangeable inside the model: on matched items with the same wrong answer, causal interventions can selectively suppress one without equally affecting the other. In three open-weight families, removing a fitted source direction lowers source compliance by 65-80 percentage points while removing a user or assistant direction has far smaller effects, and removing the user direction shows the reverse preference. A separately fitted intervention derived from source-versus-user cue activations moves compliance in both directions while leaving the
prompt text unchanged. An authority direction fitted on trivia also transfers to PIQA and multi-turn SYCON dialogues without refitting, and removing it lowers wrong-source compliance by tens of percentage points in four of five families with no detected change in MMLU-Pro or GSM8K accuracy at our evaluation sizes. Source deference and user agreement therefore need separate evaluation.
Read the full abstract ▾
Number Theory
On minima of theta and Epstein zeta functions in dimension four
Senping Luo and Juncheng Wei
Let the theta and Epstein
zeta functions be $\Theta(\alpha,L)=\sum_{v\in L}e^{-\pi\alpha|v|^2}$ for $\alpha>0$ and $E(L,s)=\sum_{v\in L\setminus\{0\}}{|v|^{-2s}}$ for $s>2$, respectively. We consider full-rank
lattices $L\subset\mathbb R^4$. Let the covolume of $L$ be one, and let $\mathcal D_4=2^{-1/4}D_4$ be the root
lattice $D_4$ rescaled to covolume one. We prove that, up to orthogonal transformations, \begin{equation}\nonumber arg\,min_{covol(L)=1}\Theta(\alpha,L)=\mathcal D_4 \qquad\text{if}\quad \alpha>0, \end{equation} and this implies that \begin{equation}\nonumber arg\,min_{covol(L)=1}E(L,s)=\mathcal D_4 \qquad\text{if}\quad s>2. \end{equation} This proves that the root
lattice minimizes the theta and Epstein
zeta functions among all
lattices in dimension four, as conjectured by Sarnak-Strömbergsson (\cite{SS}, 2006), who established the local minimality of $\mathcal D_4$. Thereby, this resolves the Rankin-Sobolev problem in dimension four dating back to Éndibaev (\cite{End1978}, 1978) and Shushbaev (\cite{Shu1978}, 1978).
Read the full abstract ▾
Computers and Society
Digital Labor Platforms as Infrastructure of Extractivism
Julian Posada
This chapter examines how digital platforms, as infrastructures of extractivism in contemporary artificial intelligence, enable corporations to appropriate data and labour while concentrating value in core regions such as Silicon Valley. It proposes platform extractivism as a framework linking commodification, epistemic dominance, and reproduction through territory. Drawing on digital labour studies, Latin American critical thought, and mixed-methods research with Venezuelan and Colombian data workers, the chapter discusses how AI depends on precarious platform labour to produce, clean, annotate, and maintain data. It further shows how annotation systems normalize culturally specific categories, especially US norms, as universal knowledge, reproducing colonial hierarchies. Finally, it highlights how data work is sustained by unpaid reproductive labour, community networks, and material conditions that make workers available for platform labour. Platform extractivism thus reveals AI as a labour-intensive and territorially embedded political-economic system under capitalism that perpetuates legacies of coloniality.
Read the full abstract ▾
Solar and Stellar Astrophysics
Abundance analysis of three new extremely metal-poor stars, including one CEMP-s star
P. François, T. Suda, S. Wanajo et al. (+5 more)
We report a detailed abundance analysis of three extremely metal-poor (EMP)
stars thanks to spectra obtained with the high-dispersion spectrograph (HDS) at the Subaru
telescope. The aim is to observe a sample of EMP candidate
stars selected from the Sloan Digital Sky Survey (SDSS) and determine their detailed chemical composition. High-resolution spectra of three faint turn-off
stars using HDS on the Subaru
telescope were acquired.We used standard 1D models to compute the abundances of several elements (Li, C, Mg, Ca, Sr, and Ba). The
stars we analysed have a
metallicity [Fe/H] of between -3.02 and -3.57 dex. We also measured accurate radial velocities that, coupled with the
Gaia astrometric data, allowed us to compute actions and
Galactic orbits for the
stars. Our analysis confirms that the
stars are genuine EMP
stars. One of them (SDSS J074238.68 +470537.0) can be classified as a CEMP-s
star with the highest [Mg/Fe] ratio ever measured in a turn-off CEMP-s
star with [Mg/Fe] = 1.85 dex. For the
star SDSS J012630.30+073029.3 ([Fe/H]= -3.02 dex), we have a clear detection of lithium with A(Li) = 2.05. Kinematically, SDSS J214340.08-002835.5 is a thick-disc
star, while the other two belong to the
halo and are candidate members of the
Gaia- Sausage-Enceladus.
Read the full abstract ▾
All authors P. François, T. Suda, S. Wanajo, E. Caffau, P. Bonifacio, W. Aoki, S. Yamada, M. Y. Fujimoto
Robotics
Cascaded consensus splitting for multi-branch contingency games
Bastien Lechardoy, Pau de las Heras Molins, Thibault Lahire et al. (+4 more)
Contingency games enable
agents to anticipate and plan for other
agents' hypothetical intents by constructing trajectories with a shared prefix and intent-dependent branches. While contingency games capture intent uncertainty, existing formulations rely on a single branching time, oversimplifying interactions in which different
agents' intentions are revealed at different times. Moreover, the computational cost of such problems grows rapidly with the number of
agents and intents, as all scenario-dependent best responses must be solved jointly. We introduce a multi-branch contingency architecture in which sources of intent uncertainty can be resolved progressively at different branching times, allowing the planned trajectories to adapt to the asynchronous revelation of intents. We also develop an ADMM-based solver that exploits scenario-level parallelism. Experiments on tightly coupled three-
agent interactions support that the proposed architecture outperforms the conventional single-branch while achieving a lower mean receding-horizon solve time.
Read the full abstract ▾
All authors Bastien Lechardoy, Pau de las Heras Molins, Thibault Lahire, Laurent Pautet, David Filliat, David Fridovich-Keil, Georgios Bakirtzis
Audio and Speech Processing
Selective Lookahead for Attention-Based Streaming ASR
Yichen Jia, Bastiaan Tamm and Hugo Van hamme
End-to-end
attention-based
speech recognition is accurate offline but hard to stream: outputs can depend on future audio, and a little future context per layer makes the lookahead grow with the number of layers. We address this with two mechanisms. A bounded-lookahead chunk encoder caps every chunk's future receptive field at a constant number of chunks, independent of the number of layers, via one age-selection rule shared by self-
attention and the depthwise convolution. On this encoder, dynamic future-chunk decoding lets a per-
token trigger commit a
token or wait and re-decode it; we propose a learned trigger as the general mechanism, with a simple confidence threshold as an effective fallback. On full LibriSpeech test-clean the dynamic system matches the best static-lookahead accuracy (6.5%) at a median latency of 306 ms versus 860 ms for one-chunk static lookahead, and a wait budget bounds the deferral tail below the static baseline's 90th percentile at 0.1 points more WER.
Read the full abstract ▾
Rings and Algebras
Schreyer Resolutions over the Exterior Algebra
Janko Boehm and Lakshmi Ramesh
Schreyer's
algorithm is usually the fastest way to determine a (typically non-minimal) free resolution of a finitely presented module over a
polynomial ring. We adapt a refined version of Schreyer's
algorithm to compute free resolutions over the exterior
algebra, relying on relative Groebner bases. We illustrate the use of our
algorithm for the computation of
cohomology of coherent sheaves over projective space, which by the BGG correspondence can be computed via free resolutions over the exterior
algebra. Schreyer's method relies on a tree traversal and thus has the potential for parallel computations. We report on ongoing work on a massively parallel implementation, observing that Gnawali's parallel approach over
polynomial rings can be carried over to our setting.
Read the full abstract ▾
Machine Learning
PHASE: Multi-Regime Modeling of Incompressible Magnetohydrodynamics
Radhika Achikanath Chirakkara, Rajdeep Haldar, Zezheng Song et al. (+1 more)
Magnetohydrodynamics (MHD) is central to
plasma modeling in astrophysics, space science, fusion, and engineering, but resolving multiscale MHD dynamics is computationally expensive. Machine-learning surrogates enable fast inference by learning reusable solution operators, yet existing models require separate training for each physical regime, limiting generalization across varying parameter settings. We introduce PHASE, a PHysics-Adaptive Scalable operator with residual Error correction, designed to model incompressible MHD across varying physical parameters with a single model. PHASE combines transfer learning, regime-aware adaptation, physics-centered learning, and residual refinement to improve both physical fidelity and generalization across MHD regimes. Together, these improvements achieve state-of-the-art prediction accuracy on two-dimensional MHD
turbulence by reducing relative $L_2$ errors on physical fields by more than an order of magnitude compared to prior MHD neural-operator baselines. Moreover, PHASE generalizes successfully to unseen parameter values without retraining, demonstrating the cross-regime adaptability expected from operator learning. We evaluate PHASE beyond point-wise prediction errors using derived physical fields, spectral analysis, and distribution statistics, consistently observing improved physical fidelity. We further show that our framework can accurately simulate MHD instabilities by testing it on the Kelvin–Helmholtz instability, demonstrating the robustness of our method.
Read the full abstract ▾
All authors Radhika Achikanath Chirakkara, Rajdeep Haldar, Zezheng Song, Jiequn Han