Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 30 August
Large language models (LLMs) can memorize sensitive information, raising serious privacy concerns. Machine unlearning offers a potential solution to remove such information, but it remains unclear whether existing methods truly erase it or merely hide it within the model. A key challenge is quantifying the persistence of sensitive data under a unified evaluation framework. To address this, we construct a synthetic dataset containing fake private information and propose a white-box auditing framework to systematically assess whether claimed-forgotten information is genuinely removed. Using this framework, we evaluate five existing unlearning methods and find that a simple "inverse greedy" decoding – selecting the least likely token at each step – can recover supposedly forgotten private information. Our results reveal that current unlearning approaches often fail to fully eliminate sensitive information, highlighting the need for more reliable methods to ensure privacy in deployed LLMs.
Cryptography and Security
Influence Is Not Authority: When Causal Guardrail Signals Make Legitimate Tool Use Look Like an Attack in Tool-Using LLM Agents
The key limitation of current state-of-the-art influence-based guardrails is that they do not reliably distinguish a legitimate, user-authorized action from a malicious, unauthorized action when both rely on external tool information. This ambiguity can cause benign actions to trigger unnecessary verification and intervention, reducing utility and adding latency. We expose this limitation through an authorization-equivalence audit of 96 conditions derived from 24 base cases. Within matched source comparisons, we hold authorization, the exact committed action, and its intended effect fixed, changing only whether a required value comes from the user or a legitimate tool result. Although the action remains unchanged, this harmless relocation shifts the causal signal toward the attack region in all 24 cases under both Llama and Gemma scorers. Matched unauthorized controls show that the signal remains attack-sensitive, yet the benign relocation produces a larger average score shift than the actual change in authorization. Architecture-level evaluation shows how this mismatch propagates through guardrail designs. With a semantic monitor, attack success is 0% and utility is 28%, compared with 16% and 60% without it. A shadow-based guardrail allows every tested harmless run, yet does not reject matched unauthorized actions more often overall: 57.5% of unauthorized runs pass automatically before reaching the later security check, compared with 29.2% of authorized runs. These results show that the studied causal signal reveals what shaped an action without reliably encoding whether the action was authorized, and that reference construction and routing are integral to the effective security decision.
Read the full abstract
Analysis of PDEs
Failure of a Brunn-Minkowski-type inequality for the Gaussian torsional rigidity
Let $u$ be the torsion function for the Ornstein-Uhlenbeck operator on a bounded domain $\Omega \subset \mathbb{R}^n$, i.e., the solution of $\Delta u - x \cdot \nabla u = -1$ in $\Omega$ with $u = 0$ on $\partial\Omega$. Let $T_\gamma(\Omega) = \int_\Omega u \, d\gamma$ be the Gaussian torsional rigidity. We prove that the Brunn-Minkowski-type inequality $T_\gamma((1-t)\Omega_0 + t\Omega_1)^{\alpha} \le (1-t) T_\gamma(\Omega_0)^{\alpha} + t T_\gamma(\Omega_1)^{\alpha}$ fails for every exponent $\alpha > 0$ and in every dimension $n \ge 2$, for a pair of convex bodies centrally symmetric with respect to the origin, which may be taken smooth with positive curvature. This answers Conjecture 1.4 for any $n \geq 2$, and Question~(Q), of Marín Sola and Salerno in the negative. The mechanism is a first-order lower bound for $T_\gamma$ at a ball $\Omega_0$ under Minkowski perturbations. When the perturbing body $\Omega_1$ has small torsion and large mean width, $T_\gamma((1-t)\Omega_0 + t\Omega_1)$ increases to first order. Since the Minkowski combination has larger torsion than both endpoints, no exponent can repair the inequality. For $n=1$, convexity with the optimal exponent $1/3$ holds on symmetric intervals by results of the same authors, \cite{MSS26}. We prove that the logarithm of the torsion is neither convex nor concave along Minkowski combinations of symmetric intervals, that no non-zero exponent yields concavity, and that convexity fails for every positive exponent when one set is a union of two intervals or when the sets are reflected off-center intervals.
Read the full abstract
Probability
Counterexamples to the site-percolation analogues of the Easo-Severo-Tassion cutset theorems
Easo, Severo and Tassion [EST25] proved two theorems about minimal edge cutsets in infinite graphs: that non-triviality of the uniform critical percolation parameter is equivalent to exponential growth of the number of minimal edge cutsets of size n separating a vertex from infinity, and that uniform transience is an alternative sufficient assumption. We show that the naive extensions of (the original contribution to) these theorems to site percolation and vertex cutsets are false; our counterexample involves modifying a rooted binary tree and replacing edges with gadgets that allow many choices for where to block (in constructing a cutset) one of two local routes, while keeping every vertex uniformly close to a supercritical, uniformly transient shortcut skeleton.
Read the full abstract
Fluid Dynamics
Data-Driven Design Optimization of Streaming-Potential-Mediated Electrokinetic Transport of Viscoelastic Fluids in Microchannels
Streaming-potential-mediated transport of viscoelastic fluids has attracted research attention owing to its applications in electrokinetic energy conversion and microfluidic transport. Existing analytical and semi-analytical models in published literature provide valuable physical insights, but require repeated numerical evaluations for exploring large design spaces and identifying the optimal operating conditions. In this work, a surrogate-assisted framework is developed for rapid design optimization of pressure-driven electrokinetic transport of simplified Phan-Thien-Tanner fluids in a slit microchannel. A high-fidelity numerical database is generated over a broad range of governing dimensionless parameters, which includes the zeta potential, the Debye parameter, the Dukhin number, and the viscoelastic parameter. A Neural Network surrogate model is subsequently trained to accurately approximate the nonlinear relationship between the governing parameters and the streaming potential, while the volumetric flow rate and hydroelectric energy conversion efficiency were calculated from closed form equation by using the streaming potential predicted by the surrogate. This is coupled with a multi-objective optimization strategy to identify operating conditions that simultaneously maximize energy conversion efficiency and volumetric flow rate. The proposed methodology can significantly accelerate parametric exploration compared with repeated numerical simulations across different parameters and provides practical design guidelines for electrokinetic microfluidic devices. The study demonstrates the potential of combining computational fluid mechanics with data-driven surrogate modeling for efficient engineering design and optimization.
Read the full abstract
Number Theory
Average twin prime conjecture for elliptic curves in arithmetic progressions
In 1988, Koblitz conjectured an asymptotic formula for the number of primes $p \le x$ for which the reduction of an elliptic curve over $\mathbb{Q}$ has prime order. Building on the work of Balog, Cojocaru, and David, who proved this conjecture on average in 2011, we extend the result to primes $p$ lying in arithmetic progressions. Our asymptotic formula holds uniformly for moduli up to a fixed power of $\log x$. We also show that the resulting average constant matches the theoretical constant predicted by Lee, Mayle, and Wang in 2025 using Galois representations.
Read the full abstract
Artificial Intelligence
AcrossWAM1.0:A Modular Latent World-Action Stack for Compact Robot Policies
Latent world-action models avoid rendering future pixels by predicting an action-relevant visual subgoal in feature space. LaWAM established this formulation, but its original presentation left the world model, multimodal backbone, and deployment checkpoint tightly coupled. We introduce AcrossWAM1.0, a modularization and scaling study of this latent world-action stack. Rather than presenting latent subgoals as a new algorithm, we make the module boundary explicit: a policy adapter produces latent-action and action-generation contexts; a retained latent world decoder grounds the predicted transition in the current scene;and a flow-matching expert generates continuous action chunks. We further separate training-only teachers from the inference graph and provide a verifiable deployment export. On 2,000 paired LIBERO episodes, replacing a Qwen3-VL-2B backbone with Qwen3.5-0.8B yields 97.45% success versus 98.00% for the 2B model (a-0.55percentage-point difference; exact McNemarp=0.266). This does not prove equivalence, but it meets a prespecified two-point retention criterion. The compact, inference-reachable checkpoint contains 1,472.6M unique parameters, 42.4% fewer than the original 2B policy, while all retained tensors are bitwise identical to the source checkpoint. Cross-family execution is additionally checked with a MiniCPM-V adapter smoke test; closed-loop cross-family transfer remains an open evaluation. AcrossWAM1.0 therefore contributes an auditable software and evaluation boundary for compact latent world-action policies, distinct from LaWAM's original latent-subgoal contribution.
Read the full abstract
Computation and Language
When Safety Speaks a Language: A Mechanistic Analysis of Safety-Language Identity Entanglement in LLMs
Safety alignment of large language models (LLMs) degrades across languages, yet the internal mechanism driving this asymmetry remains poorly understood. Our work, therefore, presents a systematic mechanistic analysis of multilingual safety using sparse autoencoder (SAE) features, sparse interpretable directions in the residual stream associated with harmful and harmless model behavior across three instruction-tuned LLMs, eight languages, and all model layers. We observe that safety-relevant features are architecture-dependent in terms of where they are located and how they are distributed across layers. Additionally, they are geometrically entangled with language identity and exhibit cross-lingual sharing patterns, i.e., languages share safety features to varying degrees across model depths and architectures. This safety-language entanglement has direct consequences such that ablating safety features impacts not only harmful response rates but also target language, with the degree of intervention predicted by the relationship between safety and language features. Our findings qualify the language-universality of safety alignment as architecture-dependent and offer a mechanistic account of multilingual safety interventions.
Read the full abstract
Robotics
System Identification of Admittance Models for Large Real-World Objects
Simulation of admittance-type models requires physically consistent dynamic models that are rarely available for off-the-shelf, everyday objects, limiting the fidelity of haptic interfaces that rely on such simulations. This paper presents the first complete workflow for producing physically consistent models of large real-world objects with various constraints and mechanisms, guaranteeing physical consistency of inertia and friction parameters. The workflow separates each object and identifies the handle and body in two stages, requiring no torque sensors at hinges, axles, or other constrained joints. Models are produced for a heavy, closer-actuated door and a wheelbarrow, representing objects of differing constraint types and model complexity. The door is modeled using four-bar linkage kinematics and a fluid dynamics-based lumped parameter model including opening, backcheck, swing, and latch zones. The wheelbarrow is modeled as a rigid body with a spherical wheel and no slip during rolling. Handle estimation RMS errors were below 0.64 N and 0.042 Nm across both objects. Door body estimation had RMS error of 2.19 Nm and wheelbarrow body estimation had RMS error of 6.77 Nm.
Read the full abstract
Computation and Language
Compression-Aware Abstention: Teaching LLMs to Refuse When KV-Compression Masks Remove Answer Evidence
KV-cache compression reduces LLM inference memory by evicting context tokens, but when the evicted tokens contain answer-bearing evidence, the model may hallucinate instead of recognizing that the compressed context is insufficient. We address this failure from a behavioral perspective: to our knowledge, this is the first work to formulate compression-aware abstention as a learning problem, in which a model learns to answer when supporting evidence survives compression and abstain when it does not. We construct supervision from compressor survival masks and tight answer-bearing spans, labeling examples as Confident when evidence survives and Abstain when it is removed. A 10.1M-parameter LoRA adapter trained on ~2.6K MuSiQue 2-hop QA examples reduces base-model hallucinations by 97% under prompt-style truncation while preserving correct answering on evidence-retaining examples. Unlike prompt-only abstention baselines, which over-abstain on many answerable high-retention examples, the trained adapter learns a conditional policy. We also evaluate the method under actual compressed-cache decoding, where multi-compressor training yields a 6-22x relative lift over the unaided base on evidence-retaining examples. Controlled-deletion experiments show that the learned behavior is driven by evidence content rather than input length alone.
Read the full abstract
Quantum Physics
Exact Entanglement Swapping through Single-Occupancy Measurements in Gaussian Fermion States
We determine the exact entanglement structure of the conditional state obtained by measuring $m$ corresponding rungs ($m\leq N/2$) in two identical copies of an arbitrary half-filled free fermion Gaussian state and post-selecting the same normalized single-fermion state ($|\psi\rangle = u|10\rangle+v|01\rangle$, where 0 and 1 denote the fermion occupancy on each sites) on each rung. For $m<N/2$, the conditional wavefunction generally depends on the initial state. Nevertheless, whenever the selected outcome has nonzero probability, the state on the unmeasured sites remains Gaussian and factorizes exactly into $N-m$ different orthogonal modes including $m$ inter-copy entangling modes and $N-2m$ spectator modes localized in one copy. Consequently, the entanglement entropy between the unmeasured parts of the two copies is $S=m h_2(|v|^2)$, independent of the initial state, where $h_2(x)=-x\ln x-(1-x)\ln(1-x)$. The success probability is given by $P_m=\det C_L\det(I_m-C_L)=\det(C_{LR}C_{RL})$, determined solely by the initial correlations and independent of ($u$,$v$). Equal-weight Bell post-selection serves as a special case that achieves the maximal entanglement swapping.
Read the full abstract
Numerical Analysis
Advancements in Spectral Collocation Methods for High-Order Eigenvalue Problems
This paper focuses on computing spectral solutions for high-order eigenvalue problems using an efficient discretization method based on Chebfun spectral discretization algorithms and domain truncation. We solve several numerical eigenvalue problems, demonstrating both the accuracy and computational efficiency of the proposed approach.
Read the full abstract
High Energy Physics (Theory)
Carrollian Wave Equations for Arbitrary Spin: Anyons and the Exotic Particle on the Noncommutative Plane
We construct the Carrollian limit of the Cortés–Plyushchay wave equations for planar anyons, grading the spin-tower components by powers of $c$ and taking the limit at fixed rest energy. The result is a linear system describing Carrollian particles of any real spin — Carrollian anyons — and, at (half-)integer spin, Carrollian bosons, fermions and higher-spin fields. The system is a first-order square root of the standard Carroll wave equation $(\partial_t^2+E^2)\Phi=0$, whose spectrum consists of two flat branches at plus and minus the rest energy. Scaling the spin along with $c$, in the Carrollian analogue of the Jackiw–Nair limit, yields instead the exotic Carroll algebra, with a second central charge, non-commuting boosts, and observable coordinates spanning a noncommutative plane. We show how the Galilean and Carrollian theories branch from one common similarity transformation: the Galilean limit expels the negative-energy modes, whereas the Carroll limit retains both relativistic branches.
Read the full abstract
Computation and Language
Token Counts Are Not Model Lineage: A Frozen-Threshold Holdout Study of Black-Box LLM API Fingerprinting
Black-box model attribution is increasingly relevant when large language models (LLMs) are served through relay and reseller APIs. A tempting low-cost signal is the prompt-token count returned by an OpenAI-compatible endpoint: two models that share a tokenizer and chat template may produce the same count sequence up to a fixed offset. Yet the validity of this signal for broader model-family attribution has received little direct holdout testing. We conduct a frozen-threshold study over 24 labeled endpoint pairs, split evenly into a development set and an untouched holdout set, with three temporal repeats and 30 controlled texts per pair. We introduce a validity-gated result contract that distinguishes an observed dissimilarity from an uninformative measurement caused by missing usage data, rate limits, or endpoint policy. The resulting shift-invariant exact-match score perfectly separates the 12 development pairs, yielding a frozen threshold of 0.725. On holdout, however, only 6 of 12 pairs are eligible under the pre-specified three-repeat rule. Among eligible pairs, balanced accuracy is 0.75, sensitivity is 0.50 (95% Wilson interval 0.15–0.85), and specificity is 1.00 (0.342–1.00). Two same-family pairs—Qwen 3.8 and DeepSeek V4 variants—fall below the frozen threshold. Across 4,320 formal API calls, every log is replayable, while holdout contains 189 non-200 responses and 157 successful responses without prompt-token usage. The study therefore validates token-count consistency as a fingerprint of a shared tokenization stack, but rejects its use as a standalone necessary test for model-family lineage.
Read the full abstract
Computer Vision
Evaluating 2D and 3D-Aware Vision Foundation Models for Vehicle Attribute Recognition
Vehicle attribute recognition is an important task in intelligent transportation systems, particularly when Automatic License Plate Recognition (ALPR) is unavailable or unreliable. Although vision foundation models have shown strong transferability across domains, their effectiveness for fine-grained vehicle classification remains underexplored. Moreover, given the inherently three-dimensional structure of vehicles, it is unclear whether emerging 3D-aware foundation models offer advantages over standard 2D architectures. This paper presents an empirical benchmark of 14 state-of-the-art 2D and 3D-aware vision foundation models. Using the challenging real-world UFPR-VeSV dataset, we evaluate these models as frozen feature extractors via linear probing for vehicle type, make, and model recognition. We further stress-test the best-performing models under few-shot learning and Out-of-Distribution (OOD) domain shifts. Our results show that standard 2D self-supervised models, particularly DINOv3, substantially outperform 3D-aware models in fine-grained tasks, achieving over 93% Macro-Accuracy for make and model recognition. However, the 3D-aware Depth Anything v2 exhibits stronger invariance to viewing angles in vehicle type classification. These findings motivate hybrid approaches that combine 2D and 3D priors for robust vehicle recognition. Our code is publicly available at https://github.com/UFPR-IPASPPR/3D-Vision-Benchmark/.
Read the full abstract
Computer Vision
Biomechanical 3D Body: Self-Supervised Distillation of Biomechanical Pose from a 3D Body Foundation Model
State-of-the-art monocular body recovery methods predict mesh vertices and angles on the corresponding kinematic tree, but their outputs lack biomechanically defined joint angles that downstream applications like clinical and biomechanical analyses require. We extend an existing foundation model, SAM-3D-Body, with an additional biomechanical prediction head that, from a single RGB image, regresses the joint angles and scales of a biomechanical model. Training this model presents a challenge, as there are limited datasets of paired images and biomechanical fits. To overcome this, we supervise biomechanical outputs with in-loop optimized targets from a Levenberg-Marquardt solver performing inverse kinematics fits against markers from the mesh predictions. This allows distilling the biomechanical head from the mesh head, even from unlabeled images. To make this work with GPU-optimized biomechanical models in MuJoCo, the entire model was implemented in JAX using Equinox. We trained this distilled output head on the publicly released SAM-3D-Body dataset. We then validated this model on biomechanical fits to two publicly available marker-based datasets, MoVi and BioCV, as well as movements from a clinical cohort captured with multiview markerless motion capture. The resulting model outperforms existing models for direct regression of biomechanics from images while only slightly underperforming the state-of-the-art monocular biomechanics method that performs more costly inference-time optimization of entire trajectories.
Read the full abstract
Computer Vision
Everybody Tracking Every Body
We address the problem of 3D body pose estimation of multiple interacting people from their egocentric views with centralized coordination. Each individual wears a camera recording egocentric video and IMU data. Processing this video with VIO SLAM provides high-quality tracking of each egocentric camera through space. The first-person view from one individual provides third-person observations of other people, although these exocentric observations are sparse, intermittent, and of highly variable reliability as both cameras and subjects move. To integrate these synchronized data streams, we propose a diffusion-based approach that fuses estimates of pose based on head motion derived from egocentric camera motion with exocentric pose observations, conditioning on both observation content and reliability. Our model is trained on a mixture of single-person motion-capture data and multi-person video in order to learn rich priors for body motion trajectories and video observation reliability. Evaluation on challenging multi-person datasets suggests our fusion approach improves over motion-only and vision-only baselines in terms of both absolute and relative pose accuracy.
Read the full abstract
Quantum Physics
Emergence of Strategic Equilibria from Transverse Field Ising Hamiltonian Dynamics
Game theory studies strategic decision-making among rational agents, and many classical games can be mapped onto interaction models such as the Ising model. Quantum game theory extends this framework by allowing players to exploit quantum superposition and entanglement. In this work, we study quantum games using an operator-based formulation derived from the transverse-field quantum Ising model. We show that the Hamiltonian-driven dynamics naturally generate entangling operator which resolve the dilemma in the game settings. This leads to a clear quantum advantage over classical outcomes. Unlike standard quantization schemes based on fixed entangling gates, the present approach enables tunable entanglement, controlled directly by physical Hamiltonian parameters, providing a hardware-relevant perspective on quantum game design.
Read the full abstract
Computer Vision
Dior: Drawing the Light of Image via Material-Decoupled Illumination Representation
Controllable image relighting is an important problem in image editing, and hand-drawn scribbles provide an intuitive interface for specifying the desired illumination. However, existing methods do not establish a consistent and effective mapping between scribble inputs and relighting results, limiting their ability to control illumination intensity, chromaticity, and complex spatial distributions. We address this limitation by introducing a material-decoupled illumination representation, termed the Lumi Map, which establishes an explicit mapping between user scribbles and the resulting illumination, thereby improving both relighting accuracy and controllability. Specifically, we use a renderer to synthesize source image-Lumi Map-relit image triplets and train the model to predict the target relighting result conditioned on the Lumi Map. To mitigate the domain gap introduced by synthetic data, we further perform reconstruction training on real relighting pairs, improving the model's generalization to real-world images. Finally, we present Dior-Light, an image relighting method controlled by hand-drawn strokes. Extensive experiments demonstrate that our method outperforms existing approaches in relighting accuracy and enables effective control over illumination intensity and chromaticity on in-the-wild images.
Read the full abstract
Computer Vision
Hallucination Mitigation for Large Vision-Language Models via Implicit Feature Stabilization
Large Vision-Language Models (LVLMs) are prone to hallucinations: they fluently describe objects, attributes, and scenes that are not in the image. We connect part of this failure to a measurable property of their representations, feature instability, where mild semantics-preserving perturbations of the input cause large changes in the learned embeddings; hallucination rates rise together with this variability. Existing stability-motivated remedies are explicit, in the sense that they intervene at inference time through latent steering or constrained decoding, and pay for it on every query. We propose implicit stabilization instead: perturbation-invariance is built into the model weights during fine-tuning, and nothing extra runs at deployment. Our framework, INFUSE, first stabilizes visual and textual representations around perturbation-averaged and ground-truth anchors, then aligns the stabilized representations across modalities with bidirectional contrastive objectives. We prove that the anchor's root-mean-square deviation from the perturbation-mean representation shrinks at rate $1/\sqrt{K}$ in the number of views, and that under a Lipschitz decoder, this bounds how much any perturbation can change the model's hallucination behavior. On LLaVA-1.5, LLaVA-1.6, and Qwen3-VL-8B-Instruct, INFUSE reduces AMBER CHAIR by 46-63% relative to each base model, improves ObjHal, MMHal, HallusionBench, and POPE, and preserves VQA-v2 and TextVQA, all with no inference-time overhead.
Read the full abstract
Computer Vision
Towards Continual Test-Time Adaptation of Vision-Language Models in Open-Vocabulary Semantic Segmentation
Open-vocabulary semantic segmentation (OVSS) relies on vision-language alignment to recognize arbitrary text-defined categories, yet this alignment is fragile under continual test-time distribution shift. Our diagnostic analysis reveals that entropy minimization drives patch-level class collapse, continual updates erode vision-language alignment, and redundant gradients from low-shift samples waste computation. We propose Diversify, Anchor, and Filter (DAF), a stabilization framework that augments entropy-based adaptation with a marginal diversity loss that resists collapse, a cross-modal anchor consistency loss that constrains feature drift relative to a frozen source model, and feature salience filtering that skips low-value backward passes to offset part of the source-anchor overhead. We evaluate on five datasets spanning natural scenes, autonomous driving, underwater imagery, and remote sensing with their corrupted variants. Across the evaluated continual shifts, DAF remains stable where entropy minimization collapses, improving mIoU by over 8 points on Pascal VOC20-C, over 9 points on LoveDA, and over 3 points on Foggy Cityscapes compared to the source model, and is robust to aggressive adaptation and learning rate choices.
Read the full abstract
Number Theory
Hecke Subalgebras and Local Newforms for the Metaplectic Double Cover of $\mathrm{SL}_2(\mathbb{Q}_p)$
Let $p$ be an odd prime, and let $\overline{\mathrm{SL}_2(\mathbb{Z}_p)}$ and $\overline{K_0(p^n)}$ denote the inverse images of $\mathrm{SL}_2(\mathbb{Z}_p)$ and the congruence subgroup $K_0(p^n)$ in the metaplectic double cover $\widetilde{\mathrm{SL}}_2(\mathbb{Q}_p)$. For $n\geq2$, we study the subalgebra of the genuine Hecke algebra $H(\widetilde{\mathrm{SL}}_2(\mathbb{Q}_p)//\overline{K_0(p^n)},\eta)$ consisting of functions supported in $\overline{\mathrm{SL}_2(\mathbb{Z}_p)}$, where $\eta$ is the genuine extension of either the trivial or the nontrivial quadratic character modulo $p$. We give an explicit basis and a presentation by generators and relations, and prove that this subalgebra is commutative of dimension $2n$. We give the multiplicity-free decomposition of $\operatorname{Ind}_{\overline{K_0(p^n)}}^{\overline{\mathrm{SL}_2(\mathbb{Z}_p)}}\eta$, determining the dimensions of its irreducible constituents and their associated Hecke eigenvalues. This Hecke subalgebra acts on the $(\overline{K_0(p^n)},\eta)$-fixed spaces in irreducible admissible genuine representations of $\widetilde{\mathrm{SL}}_2(\mathbb{Q}_p)$. We compute its action explicitly on newvectors of prescribed $\eta$-type in principal series, Steinberg, even Weil, and supercuspidal representations. For supercuspidal representations, we construct the vectors by compact induction and use Ishimoto's conductor and dimension formulas. Finally, we relate our operator $\mathcal{W}_{n-1}$ to Ishimoto's local realization of Ueda's twisting operator.
Read the full abstract
Computation and Language
Sleight of Word Benchmark: Can Language Models Notice If Their Own Output Was Tampered With?
The output of a Language Model can be tampered with while the model is writing it. A simple test can thus be constructed by evaluating the model's perception of this external perturbation. In this spirit, a simple benchmark is built in which a single word is consistently substituted with another in the generation process. We call this method Sleight of Word. Two distinct axes are measured: metrics that relate to the model's surprise, as well as an evaluation of the textual reaction for 19 different open-weight language models.
Read the full abstract
Computer Vision
Continual Test-Time Adaptation via Entropy Sensitivity-Guidance in Strict Online Setting
Test-time adaptation (TTA) promises robustness under distribution shift by updating a pretrained model on unlabeled test data, but strict online TTA with batch size one and no access to source data is especially prone to drift or collapse. We introduce Sensitivity-Guided Erasing Adaptation (SEGA), a method for strict online continual TTA (CTTA) on corruption-style streams. SEGA uses a small number of structured erasures to probe how predictive entropy changes as information is removed, and uses the resulting per-sample sensitivity trajectories to coordinate recovery and sample selection rather than relying on raw entropy or batch statistics. This yields a practical feedback signal for long-horizon batch-size-one adaptation without periodic resets or model reservoirs. In experiments on ImageNet-C, CIFAR10/100-C, and corruption-generated aquaculture streams treated as controlled corruption-style proxies, SEGA yields consistent robustness and stability gains over strong CTTA baselines while reducing backward passes through sensitivity-based gating.
Read the full abstract
Computation and Language
IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil
The rapid proliferation of LLMs has further heightened the need to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we call IndicDetect, designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, while training-free methods degrade considerably under unseen generators and adversarial attacks. The severity of these failures varies across languages, with Hindi exhibiting the largest overall degradation under adversarial perturbations. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not in their peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts.
Read the full abstract
Analysis of PDEs
A partial data coefficient identification inverse problem for a semilinear damped wave operator
This manuscript deals with a coefficient identification inverse problem for a semilinear damped wave operator in a bounded domain of $\mathbb{R}^{1+d}\ (d\geq 2)$. We establish the unique recovery of the damping coefficient, zeroth-order linear term, and the coefficient of the power-type nonlinearity from the partial Dirichlet-to-Neumann map. We investigate the corresponding uniqueness problem under the assumption that the coefficients are known in a neighborhood of the boundary, while the Neumann boundary data are prescribed only on an arbitrarily small open subset of the boundary. The analysis is largely based on the unique continuation principle, Fourier Analysis and the higher-order linearization technique.
Read the full abstract
Computer Vision
FoundYou: A Unified Model for Personalized Segmentation and Retrieval
Personalized segmentation and personalized retrieval both aim to identify the same physical object across different images. While the former localizes the object within a target image, the latter retrieves images where it appears. Despite this shared instance-level objective, the two tasks have largely evolved separately and are addressed with distinct solutions. In this work, we introduce FoundYou, a unified framework built on the observation that Segment Anything 2 (SAM 2), trained to preserve object identity across video frames, inherently captures instance-level cues. We leverage this property to match objects across independent images, enabling segmentation and retrieval to emerge as two outcomes of the same instance alignment process. This unified view unlocks new capabilities beyond traditional benchmarks, including few-shot personalized retrieval and promptable personalized segmentation with flexible prompts. Extensive experiments show consistent gains over unified and task-specific methods, including +18.4 mIoU on PerMIS and +17.8 mAP on ILIAS. Performance scales with additional references and remains robust to weaker prompts. Beyond personalization, FoundYou achieves state-of-the-art results on category-level retrieval benchmarks. Notably, our approach keeps the SAM 2-small model entirely frozen and adds only 5.9 M trainable parameters, yielding a 52 M-parameter model that is over 75x faster and 20x smaller than the only prior unified solution. Code is available at https://github.com/ga1i13o/FoundYou .
Read the full abstract
Differential Geometry
Torsion algebras of Hermitian manifolds and rigidity
We regard the Chern torsion of a Hermitian manifold as a skew-symmetric complex-bilinear product on its holomorphic tangent bundle. For a compact connected Chern–Kähler-like Hermitian manifold, we prove that this product satisfies the Jacobi identity pointwise. If, in addition, the Chern holomorphic sectional curvature is strongly quasi-positive, we show that the resulting torsion Lie algebra is nilpotent and must be abelian. Consequently, the metric is Kähler. The underlying complex manifold is therefore projective and rationally connected. For general Hermitian manifolds, we construct a metric with positive real bisectional curvature on a Hopf surface which is neither simply connected nor rationally connected.
Read the full abstract
Strongly Correlated Electrons
Symmetry-Enforced Topological Structures in Quantum Phase Diagrams
We study the topological structure of the quantum phase diagram of gapped systems by identifying the noncontractibility of loops, so-called $S^1$-families, within the gapped phase diagram in which many-body Hamiltonians can have nontrivial ground-state degeneracy. We manifest the role of symmetries in such $S^1$-family classifications by the exotic symmetry interplay: (i) nontrivial mixed anomalies, (ii) semi-direct product relation between the spontaneously broken and the unbroken symmetries, and (iii) symmetry with noninvertible operators. We find that such structures lead to the novel $S^1$-family classifications inaccessible by earlier classifications based on “independent” symmetries. Furthermore, we construct lattice realizations of these $S^1$-families and explicitly demonstrate their novel algebraic structures.
Read the full abstract
Logic
Fixed points in abstract provability structures
We study fixed points in abstract provability structures (APSs), which were introduced by Beklemishev and Shamkanov as an order-theoretic framework for studying Gödel's second incompleteness theorem (G2). For one-variable APS terms built from the provability and refutability operations $\Box$ and $\boxtimes$, we investigate the existence and uniqueness of fixed points in terms of their degree. The degree of a term is the number of occurrences of $\boxtimes$. We prove that every term of degree two has a fixed point and establish a strict hierarchy among the fixed-point properties of the terms $\boxtimes\Box^k v$. We further show that suitable levels of this hierarchy guarantee the existence and uniqueness of fixed points for arbitrary one-variable APS terms of positive degree. We also obtain several G2-like non-refutability results, some of which depend only on whether the degree is even or odd. Finally, we consider APSs based on meet-semilattices. Under an additional condition, we obtain stronger fixed-point results and show that a parametrized fixed-point property implies an abstract form of Löb's theorem.