Machine Learning
Auto-Formalizing Neuro-Symbolic Predictors
Samuele Bortolotti, Weixin Chen, Han Zhao et al. (+3 more)
Neuro-Symbolic (NeSy) predictors incorporate prior knowledge into the prediction process of neural networks, ensuring that outputs satisfy specified constraints, making them particularly suitable for high-stakes applications where compliance with domain knowledge is essential. A key bottleneck in this paradigm is the acquisition of symbolic constraints: encoding domain knowledge into logical formulas remains a manual and expert-intensive process. In this work, we investigate the extent to which auto-formalization via LLMs can systematically translate textual knowledge into symbolic knowledge that can be plugged into NeSy predictors. To this end, we introduce auto-nesy-bench, a new
benchmark for evaluating constraint formalization and its impact on downstream accuracy of NeSy predictors. Through an extensive evaluation across several domains, we find that LLMs can formalize constraints to a meaningful extent, generating formulas that are often similar to those provided by human experts. Moreover, when the generated formulas are syntactically valid, they can lead to high-quality downstream predictions. The code and
benchmark are available at https://unitn-sml.github.io/auto-nesy-bench/.
Read the full abstract ▾
All authors Samuele Bortolotti, Weixin Chen, Han Zhao, Andrea Passerini, Stefano Teso, Antonio Vergari
Human-Computer Interaction
Who Thinks First? Designing Productive Friction with Engage-to-Unlock GenAI
Xiaotian Su, Laura Rimell, Jiazheng Li et al. (+6 more)
Generative AI can support writing, but frictionless access may cause cognitive offloading before users develop their own ideas. We introduce Engage-to-Unlock, a productive-friction mechanism that unlocks generative capabilities after users meaningfully engage with the task. In a controlled experiment (N = 398), participants completed a writing task under one of four conditions: Human-Only, Standard Chatbot, Engage-to-Unlock, or Time-Matched Unlock, which matched unlock timing to Engage-to-Unlock participants but independent of users' engagement, then evaluated passages for evidence and inferential errors. Results show that Engage-to-Unlock redistributed effort across tasks: participants spent more time writing and less time evaluating, without increasing overall task duration. They also submitted more
prompts than in other AI-assisted conditions and showed the highest accuracy-per-time evaluation efficiency across conditions. These findings suggest that designing GenAI access to encourage early human engagement may provide a productive form of friction, while retaining active AI use and efficient downstream evaluation.
Read the full abstract ▾
All authors Xiaotian Su, Laura Rimell, Jiazheng Li, Amal Rannen-Triki, Ulrich Paquet, Lisa Anne Hendricks, Rida Qadri, Daphne Ippolito, Piotr Mirowski
Computer Vision
SuperMotion: Source-Preserving Denoising for Text-Driven Human Motion Editing
Fa-Ting Hong and Peter Wonka
Text-driven human motion editing aims to realize a requested change while preserving compatible source content. Existing
diffusion editors rely largely on learned conditioning for preservation of the unedited part, yet their outputs can lose temporal detail as denoising proceeds. We propose the Source-Preserving Denoising framework (SuperMotion), which explicitly reuses the source at each reverse step for source preservation. We first align the source motion to the output timeline and predict a preservation gate that controls reuse across frames and feature dimensions. A clean-space source anchor then utilizes the learned preservation gate to blend the predicted clean motion with the aligned source and passes the corrected estimate directly to the sampling posterior. Because the aligned source is a realized motion rather than a regression output, the anchor injects sample-level temporal detail that a reconstruction-trained denoiser tends to smooth away. To learn effective source reuse, we supervise the anchored estimate against the editing target and match its second temporal differences through a temporal high-frequency loss. These objectives require no explicit edit masks. Extensive experiments show that SuperMotion improves editing accuracy, reaching 33.20% full-pool R@1 on MotionFix, while reducing temporal-detail attenuation and preserving motion dynamics as it realizes the requested changes. Ablations confirm that the learned preservation gate is responsible for the gain and that it reuses the source to retain the unedited content properly.
Read the full abstract ▾
Combinatorics
Tight degeneracy bounds in online Ramsey games
Wen Chen, Qizhong Lin and Shixi Song
In the $q$-color online Ramsey game, Builder and Painter play on an infinite independent set of vertices. At each step, Builder draws an edge and Painter immediately assigns it one of $q$ colors. Builder aims to force a monochromatic copy of a fixed graph $H$. We prove that, for every $q \ge 2$ and $d \ge 1$, Builder can force a monochromatic copy of any $d$-degenerate graph $H$ while drawing a graph of degeneracy at most $d$. The bound $d$ is tight, and this resolves in the affirmative a problem of Conlon, Fox and Sudakov.
Read the full abstract ▾
Machine Learning
FedMIX-P: Mixing Local and Global Preconditioners for Federated Vision and Language Model Training
Junkang Liu
Adaptive preconditioners accelerate model training, but heterogeneous client geometries can bias
federated updates even when gradients are evaluated at the same model. Round-start synchronization alone cannot prevent this mismatch from reappearing during local training. We propose FedMIX-P, which mixes shared and local preconditioners at every local step, retaining local adaptation while reducing mean-squared operator mismatch by a factor of $\lambda^2$. For smooth nonconvex objectives with
stochastic gradients and partial participation, we establish an $O(R^{-1/2})$ stationarity bound using suitable stepsizes and a horizon-dependent mixing weight, without requiring local preconditioners to converge to one another. A two-client counterexample shows that fixed positive mixing can preserve a nonstationary fixed point. The theory covers bounded linear symmetric positive-definite preconditioners. Experiments with SOAP, Sophia, and
Muon variants across
vision and
language tasks show improvements over corresponding local optimizers, including accuracy gains of up to $19.47$ percentage points and lower validation loss for 60M–350M
language models. Full nonlinear and momentum-based updates require separate analysis.
Read the full abstract ▾
Computation and Language
How the Audit Rule Shapes Faithful Factor Explanations in LLMs
Taolin Zhang, Hanyu Wang, Jiuheng Wan et al. (+2 more)
Large
language models are often asked which input factors influenced their outputs. For structured inputs, such reports can be checked by counterfactual perturbation, but each factor must be queried multiple times to estimate its effect, so verification is usually budget-limited. We study how this limited-budget setting changes the incentive to report factor-level influence truthfully. We formalize the interaction as a verification game and show that proper scoring alone is not enough when
auditing depends on the report: report-dependent
auditing creates a suppression incentive, because factors reported as important are more likely to be checked and penalized for estimation noise. In contrast, report-independent
auditing, or a mixed rule with a small report-independent floor, removes this channel and makes truthful reporting preferable to full suppression. We instantiate the framework with the Counterfactual Brier Score (CBS) and evaluate its predictions on four NLP
benchmarks. A synthetic rational
agent matches the theoretical prediction exactly, and real LLMs follow the same incentives when they are made explicit. The main design implication is simple: under partial verification, factor-level explanation systems should include a report-independent audit component so that under-reporting cannot be used to avoid scrutiny.
Read the full abstract ▾
All authors Taolin Zhang, Hanyu Wang, Jiuheng Wan, Tingyuan Hu, Chengyu Wang
Artificial Intelligence
Decision Titan: Test-Time Training for Long-Term Memory in Offline Reinforcement Learning
Jude Waide and Robert Lieck
Long-term dependencies remain a major challenge for sequential decision-making in the field of AI: RNNs suffer from vanishing gradients and the limited expressivity of vector-based hidden states, whilst
Transformer-based models are limited by the quadratic scaling of
attention. Recent work has proposed tackling this problem with the Test-Time Training (TTT) framework, which stores episodic
memories in the parameters of a neural network through gradient descent at both train and test-time. This approach has seen success in the domain of Natural
Language Processing, however, to the best of our knowledge it has not yet been applied to the domain of Reinforcement Learning (RL), nor has there been a study analysing how this
memory practically functions. In this paper, we study the potential of the TTT framework for offline RL by augmenting a Decision
Transformer with TTT layers, dubbed the Decision Titan. We analyse performance and properties of the model in the X-Maze environment, an extension of T-Maze designed to test sequential
memory, and investigate how the
memory mechanism learns by visualising gate values over time. Our key findings are that Decision Titan can learn long-term dependencies with ranges 20x longer than the context window, generalises to lengths 1.7x the training data, but crucially temporal generalisation depends on the time
embeddings used, and the ability to learn long-term dependencies depends on how the relevant information is encoded.
Read the full abstract ▾
Computer Vision
VoxelSynth3D: Interpretable Volumetric Image-Domain Metal Artifact Reduction with a Paired Synthetic CLINIC Benchmark
Amritesh Banerjee, Abdul Basit, Renil Renji Joseph et al. (+2 more)
Metal artifacts in postoperative musculoskeletal CT obscure bone-implant and adjacent soft-tissue interfaces. Many metal artifact reduction (MAR) methods require unavailable raw projections or learned models that may shift across scanners and implants. We present VoxelSynth3D, a training-free 3D image-domain framework for reconstructed CT. The framework combines support masking, normalized tissue synthesis, deviation gating, and restricted edge refinement. Detected implant voxels are preserved in the output, while correction targets metal-induced artifacts in the surrounding tissue. We also construct Synthetic CLINIC-Metal, a controlled paired synthetic evaluation resource, from no-metal CTPelvic1K volumes with clean targets, metal/artifact masks, fixed seeds, and patient-level splits; 75 unpaired real metal cases receive qualitative/no-reference evaluation only. The operating point was fixed in a near-flat validation basin. With exact-mask oracle localization, all methods share a metal-excluded tissue ROI. On 40 held-out cases, VoxelSynth3D reduced RMSE from 801.48 to 786.18 HU (paired gain 15.30 HU, 95% CI 11.68-19.23), improving every case and exceeding the evaluated 3D Gaussian smoother by 13.58 HU. Clean-edge agreement decreased next to metal but exceeded input beyond 5 mm. Thus, VoxelSynth3D provides case-consistent within-distribution tissue-error reduction with a localized structural tradeoff. Spacing-aware sensitivity retained aggregate broad-region improvement and identified near-metal calibration as a target.
Read the full abstract ▾
All authors Amritesh Banerjee, Abdul Basit, Renil Renji Joseph, Nouhaila Innan, Muhammad Shafique
Computation and Language
GAW-PO: Preference Optimization with Gradient-Aligned Token Weights
Andreea Dutulescu, Stefan Ruseti, Mihai Masala et al. (+2 more)
Most preference optimization methods, such as Direct Preference Optimization (DPO), apply preference supervision at the response level, although autoregressive
language models are optimized
token by
token. As a result, all
tokens in a rejected response contribute to the negative training signal, including
tokens that may encode behavior that is useful for the preferred response. We introduce GAW-PO, a gradient-aligned
token reweighting method for DPO that estimates, for each rejected
token, whether penalizing it would interfere with the preferred update directions.
Tokens whose gradients are strongly aligned with the preferred behavior receive a weaker negative contribution, while conflicting
tokens retain a stronger penalty. Our method achieves the highest average performance among the evaluated preference-optimization methods, improving by 0.97 points over standard DPO and 0.65 points over the strongest competing baseline across 11
benchmarks spanning mathematics,
reasoning, coding, and question answering. We further show that gradient-aligned weighting is substantially more robust to aggressive preference optimization: as the DPO regularization parameter $\beta$ decreases, standard DPO degrades sharply, whereas GAW-PO continues to improve. These results suggest that accounting for the interaction between rejected-
token updates and preferred behavior provides an effective form of
token-level credit assignment for preference optimization.
Read the full abstract ▾
All authors Andreea Dutulescu, Stefan Ruseti, Mihai Masala, Traian Rebedea, Mihai Dascalu
Computer Vision
FedCKA: Representation-Guided Layer Personalization for Federated 3D Perception Across Driving Domains
Jolle Verhoog, Ali Burak Ünal and Holger Caesar
Robust perception in intelligent vehicles demands 3D object detectors that remain dependable under domain shifts, such as changes in time of day, location, or weather. However, due to costly annotation and rare shifts, some environments lack sufficient data to train a standalone detector.
Federated learning offers a
privacy-preserving framework for collaborative model training, enabling clients to benefit from shared learning across diverse environments. Yet, this framework traditionally relies on a single global consensus model, which struggles to perform across heterogeneous local data distributions. Local conditions are better captured by adapting a subset of the model, but many personalization approaches rely on predefined layer partitions or fixed personalization ratios, thereby limiting adaptation to client-specific divergence. To reduce this
rigidity, we propose FedCKA, a Centered Kernel
Alignment (CKA)-based strategy that dynamically handles the personalization-globalization trade-off. Specifically, FedCKA computes layer-wise feature similarities between local client models and the global consensus model during training. By converting layer-wise similarity scores into client-specific aggregation masks, FedCKA selectively shares representation-consistent layers. Evaluation on a unified multi-domain
benchmark based on nuScenes shows that FedCKA outperforms established
federated baselines, including FedBN, FedRep, and FedSelect, improving average NDS by 7 percentage points over the strongest baseline. The findings offer both a comparative
benchmark and a promising direction for robust
federated 3D perception across shifts in location, weather, and illumination. Code is available at https://github.com/j-verhoog/FedCKA.
Read the full abstract ▾
Artificial Intelligence
Sharpening Tax in Post-Training
Changdae Oh, Qi Zeng, Qi Qi et al. (+7 more)
An emerging hypothesis about reinforcement learning (RL) post-training of large
language models (LLMs) is that it merely sharpens existing behaviors of a base model, improving single-shot accuracy at the cost of solution coverage. Although this trade-off has been observed in math and coding tasks, it need not extend to agentic tasks, where multi-turn tool use and interaction may require capabilities newly acquired during post-training. Our surprising finding is that pre-trained LLMs, equipped with a light inference harness, can serve as capable
agents. Despite far lower accuracy (pass@1), they often surpass their post-trained counterparts in solution coverage (pass@K) given a sufficient test-time budget. We further analyze the underlying mechanism and show that post-training pushes tasks toward two extremes, always solved or never solved, and thereby improves sampling efficiency and consistency at the cost of solution coverage. To measure this cost, we propose Sharpening Tax, a diagnostic metric that quantifies the loss in test-time scalability after post-training. Across 14 base/post-trained model pairs from four families and three agentic
benchmarks (42 cases in total), the tax is prevalent in most settings, can be estimated from a few rollouts, and correlates well with other metrics. Finally, we present posterior-tempered group sampling (PTGS), a simple plug-and-play
Bayesian sampler that adapts the sampling temperature per
prompt to its estimated difficulty. Applied during RL training in two agentic environments, PTGS pays a smaller tax than the fixed-temperature baseline, solving more tasks under repeated sampling while also improving single-shot accuracy.
Read the full abstract ▾
All authors Changdae Oh, Qi Zeng, Qi Qi, Andrey Zhmoginov, Deren Lei, Yun He, Hoang Phan, Hangoo Kang, Azalia Mirhoseini, Sharon Li
Cryptography and Security
OverAct: Measuring and Mitigating Proactive Over-Authorization in LLM Tool-Calling Agents
Taolin Zhang, Jiuheng Wan, Hanyu Wang et al. (+2 more)
LLM
agents with tool-calling capabilities can access external services and private user data, but they may retrieve more information than a user's request explicitly requires. We study this behavior in structured tool-calling
agents and term it proactive over-authorization. This setting differs from filesystem-level coding
agents because the main risk is unnecessary access to private data. We introduce OverAct, a controlled
benchmark spanning eight
privacy-sensitive domains with deterministic, judge-free scoring, together with an interpretive decision-theoretic framework that yields three testable predictions. Across seven models from four families, all models significantly exceed authorized scope. Request specificity is the strongest predictor of severity, over-authorization grows sublinearly with tool-pool size, and decoding temperature has little effect. These patterns are consistent with a cost-asymmetry account, suggesting that over-authorization arises more from structural decision tendencies than from decoding randomness. We also propose SelfAudit, a zero-shot inference-time method that generates request-grounded justifications and filters unjustified calls before execution. Ablation shows that explicit filtering is the main driver of scope reduction. SelfAudit reduces
privacy-oriented excess by 43% without oracle knowledge.
Read the full abstract ▾
All authors Taolin Zhang, Jiuheng Wan, Hanyu Wang, Tingyuan Hu, Chengyu Wang
Chaotic Dynamics
Evidence for an Obstruction to Perturbative Renormalization Group in Stochastic Turbulence
L. Ts. Adzhemyan, D. A. Davletbaeva, D. A. Evdokimov et al. (+1 more)
We examine whether the perturbative renormalization group (RG) regime of
stochastic turbulence can be continued from small $\varepsilon$ to the physical value $\varepsilon_{\mathrm{phys}}=2$, corresponding to large-scale forcing. Exploiting the simplifications of the large-$d$ limit, we calculate the correction exponent $\omega$ to five-loop order and find a simple structure suggesting a closed analytic form. This form implies that the fixed point stable near $\varepsilon=0$ ceases to be simultaneously real and infrared stable before reaching $\varepsilon_{\mathrm{phys}}=2$, pointing to a possible change of infrared regime rather than a controlled perturbative RG extrapolation.
Read the full abstract ▾
All authors L. Ts. Adzhemyan, D. A. Davletbaeva, D. A. Evdokimov, M. V. Kompaniets
Artificial Intelligence
MCRI: A Four-Dimensional Framework for Analyzing and Evaluating Agent Skills
Zongrui Yang, Li Xintong, Runchen Xu et al. (+4 more)
As
agents evolve from single-tool systems into modular, composite architectures, skills are becoming an important mechanism for capability development and distribution. However, the academic community lacks a structured framework for systematically analyzing and evaluating skills. Drawing on information gain and behavioral constraint, we propose the four-dimensional MCRI Framework and operationalize it as MCRI-Eval, a large
language model-based evaluation method. We evaluate MCRI-Eval using 63,812 public skills from the OpenClaw skill Hub, with 58,275 skill-conditioned model executions across BigCodeBench, BFCL-Fundamental, and Mind2Web. MCRI-Eval scores are positively associated with community popularity signals and achieve the highest downstream ranking agreement among the evaluated methods. MCRI-Eval also improves top-1 skill selection across all three
benchmarks: compared with the strongest baseline on each
benchmark, the skills selected by MCRI-Eval advance by 17.7, 22.8, and 19.6 percentile points in downstream performance rank on BigCodeBench, BFCL-Fundamental, and Mind2Web, respectively. These results indicate that MCRI-Eval provides a useful pre-execution signal for prioritizing promising skills before costly execution-based evaluation.
Read the full abstract ▾
All authors Zongrui Yang, Li Xintong, Runchen Xu, Zhongsheng Wang, Zhedong Lin, Haoyuan Li, Jiamou Liu
Robotics
OpenSpace Lab Solution to the IROS 2026 Indoor Exploration Competition
Yuxuan Zhang, Dong Li, Zezhou Sun et al. (+5 more)
This report presents the OpenSpace Lab's solution to the Competition on Intelligent Information Gathering for Single and Multi-
Robot Systems Workshops, organized as part of the IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS) 2026. Our team reached 1st place in the Single-
Robot Public Track and 3rd place in both the Single- and Multi-
Robot Private Tracks. The single-
robot framework utilizes pre-trained map completion predictions for global planning to prioritize unexplored areas. To reconcile map coverage with limited operation time, we introduce a remaining-time-based exploration strategy that integrates homing constraints into the decision-making process. For multi-
robot exploration, we utilize a utility-driven target selection strategy that balances observation gains, movement costs, and budget constraints, leveraging shared map and intent data to eliminate redundant search and maximize coordination efficiency. Our solution reached a 61.04% coverage rate in the Single-
Robot Public Track, while reaching 39.53% and 39.91% coverage in the Single- and Multi-
Robot Private Tracks, respectively. An extended full-length paper based on this report is currently being prepared for submission, and the source code will be released upon acceptance of the full manuscript at https://github.com/OpenSpace-Lab/Indoor-Exploration-IROS2026.
Read the full abstract ▾
All authors Yuxuan Zhang, Dong Li, Zezhou Sun, Yuxuan Xu, Siyu Teng, Yuchen Li, Jianjian Yang, Long Chen
Numerical Analysis
An Unfitted Hybrid High-Order Method for the Elastodynamics Problem with Imperfect Interface
Peiqi Huang and Erik Burman
We design and analyse an unfitted hybrid high-order (HHO) method for the elastic wave equation in a medium made of two components separated by an imperfect interface of linear slip type, across which the traction is continuous and the displacement jump is proportional to the traction through a compliancy tensor $\bK=\alpha\bI+(\beta-\alpha)\bn\otimes\bn$. The mesh is not fitted to the interface: the discrete unknowns are doubled in the cut
cells, the small cuts are cured by a
cell agglomeration procedure, and no unknown is attached to the interface. The two specific ingredients of the method are a local symmetric strain reconstruction in each cut subcell, which incorporates the interface condition through the regularised interface stiffness $\bS_h=(h_T\delta^{-1}\bI+\bK)^{-1}$ in the spirit of Hansbo and Hansbo {\em{A finite element method for the simulation of strong and weak discontinuities in solid mechanics.}} {Comput. Methods Appl. Mech. Engrg.}, 193, 2004, and an interface stabilisation built from the same
matrix. For the space semi-discrete problem we prove that the discrete bilinear form is coercive and continuous, and we derive an energy-error estimate of order $h^{k+1}$ and an $L^2$-error estimate of order $h^{k+2}$, with constants independent of the compliancy parameters and of how the interface cuts the mesh. The scheme is combined either with the Newmark scheme, which conserves a discrete energy exactly, or with singly diagonally implicit Runge–Kutta schemes of order up to four. Numerical experiments in two dimensions confirm the predicted convergence rates for $k\in\{1,2,3\}$, the robustness with respect to the compliancy over sixteen orders of magnitude, and illustrate the propagation of elastic waves across an unresolved slipping interface.
Read the full abstract ▾
Optimization and Control
McADMM: A Multi-Clique Augmented Lagrangian-Based Algorithm for Large-Scale Sparse SDPs with Bound Constraints
Kristo Nugraha Lian, Nehal Ahmed Shaikh, Di Hou et al. (+2 more)
sGS-PADMM [21, 14, 6] is a powerful and versatile class of convergent multi-block ADMM solvers for implementations on moderate-sized linear semidefinite programming (SDP) problems. In this paper, we further enhance this class of
algorithms for solving SDP problems by proposing a new multi-clique decomposition approach, allowing substantial improvements in applications on large-scale sparse SDPs (e.g., where $n > 1000$) with conducive aggregate sparsity patterns. Our SDP decomposition strategy mainly aims to reduce the estimated PSD projection cost after decomposition, in contrast to common decomposition
algorithms that are encumbered with minimizing the overlaps between cliques. This feature is made possible by our novel linear-space projection approach that is capable of efficiently processing a large number of overlap constraints via simple averaging steps. For the numerical experiments, we demonstrate the performance of our solver – named McADMM for Multi-clique ADMM – on a number of large-scale SDP instances that arise from relaxations of some important quadratically constrained quadratic programming (QCQP) problems. The performance of McADMM is contrasted against other state-of-the-art decomposition-based solvers as well as the non-decomposed sGS-PADMM to highlight our key contributions. We additionally develop a GPU implementation of McADMM and demonstrate that it can substantially accelerate the decomposed solver.
Read the full abstract ▾
All authors Kristo Nugraha Lian, Nehal Ahmed Shaikh, Di Hou, Xingyu Xie, Kim-Chuan Toh
Machine Learning
Learned End-to-End Guidance Schedules for Diffusion Models
Aneesh Barthakur, Mathias Niepert and Luiz F.O. Chamon
Diffusion models are a powerful generative paradigm used across multimedia and scientific applications. Guided
diffusion methods impose requirements on the generation by adding the gradient of a differentiable loss (the guidance function) as a drift term during inference. The weight of this drift (the guidance scale) is critical for the trade-off between data quality and requirement satisfaction. To achieve both of these goals, guided
diffusion must resort to small guidance scales and lengthy sampling, incurring high computational costs. This work proposes learned end-to-end guidance schedules (LEEGS) to achieve these objectives with fewer sampling steps. LEEGS trains a time-dependent schedule by minimizing the guidance function over a small set of examples using
stochastic gradient descent. Backpropagating through guided sampling is computationally expensive, so LEEGS uses an approximation of the gradient that cuts training time by a factor of 4. We evaluate LEEGS on diverse guidance tasks, including (a) image inpainting, (b) noisy image inverse problems, (c) face-ID-guided generation, and (d) forward and inverse PDE problems, outperforming baselines at equal budget (50 or 100 NFEs), or matching constant guidance with only 10% of the steps.
Read the full abstract ▾
Numerical Analysis
Finite element exterior calculus for spectra and pseudospectra of advection-diffusion of differential forms
Daniele Boffi, Kaibo Hu, Yizhou Liang et al. (+1 more)
Numerical investigations of dynamo action have remained active in fluid mechanics over the past decades, presenting numerous challenges and open problems. Meanwhile, the development of structure-preserving methods and finite element exterior calculus (FEEC) inspires a revisit of numerical dynamo studies and an exploration of existing open questions in computation. In this paper, we present a FEEC approach for dynamo problems. In particular, we investigate structure-preserving finite element schemes for computing the spectra and pseudospectra of advection-
diffusion operators of differential forms. The schemes and their analysis are based on finite element de Rham complexes.
Read the full abstract ▾
All authors Daniele Boffi, Kaibo Hu, Yizhou Liang, Umberto Zerbinati
Functional Analysis
Sparse Kahane–Salem–Zygmund Forms and Weighted Hardy–Littlewood Inequalities Across the Critical Endpoint
Anderson Barbosa, Daniel Núñez-Alarcón, Anselmo Raposo et al. (+1 more)
We study sparse Kahane–Salem–Zygmund constructions and weighted Hardy–Littlewood inequalities for homogeneous
polynomials. For supports of cardinality $n^{d+o(1)}$, we determine the sharp power of $n$ governing the smallest norm of a unimodular $m$-linear form on $\ell_{p_1}^n\times\cdots\times\ell_{p_m}^n$; in the diagonal case, this yields the missing
polynomial growth exponent in the coefficient-versus-supremum norm problem for $2\le p\le m$ and $2\le r\le\infty$. We then introduce a diagonal weighted Hardy–Littlewood functional which, on $s=p\ge m$, agrees exactly with the classical Hardy–Littlewood coefficient norm with the same optimal constant. We determine the optimal diagonal weight exponent for $2\le p\le m$ and $1\le q\le2$, on the full critical line $p=m$, and on a sharp part of the region $q>2$; at $q=\infty$ the optimal weight exponent is obtained for every $2\le p\le m$. The sparse coefficient estimates provide the matching dimensional obstructions.
Read the full abstract ▾
All authors Anderson Barbosa, Daniel Núñez-Alarcón, Anselmo Raposo, Joedson Santos
Computer Vision
VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation
Yu Huang, Jungang Li, Zhiyuan Wang et al. (+8 more)
Recent
video generation models can produce highly realistic
videos from natural
language instructions, with visual quality approaching cinematic standards. Existing evaluation
benchmarks, however, predominantly assess visual quality, aesthetic appeal and physical plausibility, while paying limited
attention to text, an essential medium for conveying information in everyday scenes. A generated
video may appear visually compelling and feature lifelike subjects, yet still render the text within the scene incorrectly. To address this overlooked dimension, we introduce VTR-Bench, a systematic
benchmark for evaluating the Visual Text Rendering capabilities of
video generation models. VTR-Bench situates text within concrete application scenarios, such as advertisements and scientific
videos, with 300 carefully constructed
prompts spanning five scenario categories. We develop an automated evaluation pipeline with human
alignments that separately assesses text fidelity through carrier-specific transcription and scene and motion requirements through a
prompt-specific chain of query. Beyond evaluation, we introduce a Keyframe-Guided Agentic Framework in which a Director
agent coordinates image and
video generation with visual evaluation, guiding iterative refinement and candidate selection through visual feedback. Experiments on 11 state-of-the-art models reveal widespread difficulties in accurately rendering scene text, with the best-performing model recording an overall word error rate (WER) of 0.250. We further analyze text rendering failures to characterize the challenges faced by current
video generation models. These findings highlight visual text rendering as a key challenge for
video generation and demonstrate a practical path toward improvement. Code is available at https://github.com/hardenyu21/VTR-Bench.
Read the full abstract ▾
All authors Yu Huang, Jungang Li, Zhiyuan Wang, Yonghua Hei, Song Dai, Jiayu Yang, Deyuan Liu, Xiang Zheng, Xiaoshuang Shi, Hao Cheng, Kaidi Xu
Spectral Theory
Eigenvalues of Hermitian Toeplitz matrices with Fisher–Hartwig symbols
Jacek Wszoła
We investigate the
eigenvalues of Hermitian Toeplitz
matrices generated by the symmetric Fisher-Hartwig symbol $\hat a(z) = (1-z)^{\alpha/2}(1-z^{-1})^{\alpha/2}$ for $\alpha \in (0,2)$. Using Dirichlet-Neumann bracketing of the discrete
Laplacian, we establish explicit, non-asymptotic bounds for the individual
eigenvalues. As a consequence, we prove that all
eigenvalues are simple. We also obtain a two-term approximation for every
eigenvalue, with explicit bounds on the remainder, as the size of the
matrix tends to infinity. While known results are limited to $\alpha > 1$, we bridge this gap by covering the full range of $\alpha \in (0,2)$. Our approach uses a construction of approximate eigenvectors that has not previously been applied in this setting.
Read the full abstract ▾
Artificial Intelligence
OpenMTB-Audit: Exposing Over-Refusal and Clinical Expert Perspectives in LLM-Based Molecular Tumor Board Safety Evaluation
Negin Ashrafi, Jia Luo, Stacey M. Frumm et al. (+1 more)
Molecular tumor boards integrate genomic findings, clinical context, and therapeutic evidence to support precision oncology. As AI enters this workflow, a key safety challenge is distinguishing truly unsupported recommendations from evidence-supported options that still require oncologist review because of incomplete information, poor ECOG performance status, or other clinical caveats. We introduce OpenMTB-Audit, an open-source
benchmark of 500 synthetic non-small
cell lung cancer cases spanning five
adversarial error categories and four safety labels: Supported, Partially Supported, Unsupported, and Insufficient Information. Across eight large
language model configurations, we identify pervasive over-refusal: all LLM configurations failed to retain the Partially Supported label in 83.3-100% of true Partially Supported cases, achieving high aggregate safety scores through label collapse rather than clinically calibrated
reasoning. To address this limitation, we developed MTB-AuditAgent, a deterministic seven-module framework separating evidence verification, missing-information detection, safety classification, and abstention. It reduces over-refusal to 6.7% and achieves 91.2% accuracy (95% CI: 88.6-93.6%). A two-oncologist annotation study found disagreement concentrated at the boundary between information sufficiency and treatment optimization, underscoring the need to preserve clinically meaningful distinctions.
Read the full abstract ▾
All authors Negin Ashrafi, Jia Luo, Stacey M. Frumm, Roxana Daneshjou
Computer Vision
SALD: Self-Referenced Advantage Learning for Diffusion Models
Aryan Das, Surjo Dey, Koushik Biswas et al. (+4 more)
Recent work on
language-model adaptation has shown that single models can obtain informative training signals by evaluating their behavior in demonstrationor feedback-augmented contexts, with the help of a teacher network, which is driven by the student's learned parameters. Inspired by this internal-reference principle, we investigate how
diffusion models can identify self-referenced training signals without external demonstrations or teacher networks. We introduce SALD, a self-referenced training framework that evaluates each image-caption pair at two noise levels using the same model. The easier, lower-noise path is evaluated without gradient tracking to provide a reference, while the harder, higher-noise path provides the training gradient. Rather than directly distilling the easy-path prediction, SALD uses the difference between two path errors to adapt the hardpath objective. The proposed Advantage-Guided
Diffusion (AGD) converts this relative error into a differentiable sample-level weight. Temporal Advantage
Memory (TAM) accumulates relative difficulty across training and adapts the future gap between the two noise levels. Spectral Advantage Decomposition (SAD) further compares the residual power spectra of the two paths and constructs a differentiable, frequency-derived latent-element weight. All components share a single set of model parameters, requiring neither an external teacher network nor additional trainable parameters during training or inference, and no modification to the inference procedure. Experiments across multiple architectures and
datasets demonstrate consistent improvements in generation quality, while component-wise ablations quantify the contributions of the proposed components.
Read the full abstract ▾
All authors Aryan Das, Surjo Dey, Koushik Biswas, Swalpa Kumar Roy, Moloud Abdar, Arnab Bhattacharya, Vinay Kumar Verma
Artificial Intelligence
Wenhao Liang, Lin Yue, Wei Emma Zhang et al. (+3 more)
Dynamic architectures leave a per-example routing trace beside each prediction, and diffuse routing is easy to read as a sign that the prediction is unreliable. We audit that reading for routing
entropy in
Attention-Residual (AR) variants of Swin-Tiny and DeiT-Small, trained from scratch on CIFAR-10/100 with a soft-binned calibration auxiliary loss, asking whether the trace carries information about correctness beyond what the model's own confidence already reveals. Three checks probe this increment: does a routing signal appear at fixed confidence, does it replicate across training seeds, and can a held-out predictor exploit it against output-only and shuffled-trace controls? A sensitivity audit then injects effects of known size and measures the fraction of each that the probes recover. No test in the fixed 30-test binned family survives multiplicity correction, and neither the nominal hit nor a borderline result recurs in its sibling seeds. Across 24 paired runs a scalar routing probe yields no pooled improvement in routing-stratified calibration, and an
entropy-profile probe predicts correctness better than the same probe given shuffled profiles yet worse than a confidence-only predictor in both binary log-loss and Brier score: a gain over shuffled traces does not become a gain over the output. Conditioning on the complete logit vector leaves the corresponding comparison unresolved. The audit bounds how far these non-detections can be read: at an injected effect of 0.010 nats the profile probe recovers 24-59% of the oracle gain, and a reference-preserving correction probe recovers 8% and 23% in the two CIFAR-100 settings, below the threshold we fixed for applying it to real labels. The results establish control-dependent gains and incomplete
estimator recovery, not the absence of conditional routing information.
Read the full abstract ▾
All authors Wenhao Liang, Lin Yue, Wei Emma Zhang, Mingyu Guo, Olaf Maennel, Weitong Chen
Machine Learning
Let the Heads Talk: Beyond Diagonal Graph Attention
Riccardo Ali, Alessio Borgi, Mario Severino et al. (+4 more)
Sheaf Neural Networks generalize scalar-weighted message passing by replacing scalar edge weights with linear transport maps between local feature spaces. Yet the role of this
matrix-valued transport is entangled with the broader sheaf-
diffusion construction. We isolate the transport primitive through quiver representations and establish a direct connection with multi-head
attention. Treating
attention heads as coordinates of a local transport space reveals that standard multi-head
attention implements diagonal edge maps: along each directed interaction, a source head can contribute only to the corresponding receiver head. Allowing off-diagonal entries instead enables edge-conditioned communication across heads before neighborhood aggregation. We show that this operation cannot, in general, be absorbed into a single shared linear map applied after aggregation. Building on this characterization, we introduce Topological
Attention (Top-A), a multi-head
attention that learns edge-dependent off-diagonal routes while preserving the original same-head paths and exactly recovering vanilla
attention when the additional routing vanishes. We evaluate Top-A on relational
reasoning, heterogeneous graph learning, and algorithmic
reasoning, including out-of-distribution generalization, with heterophilic node classification as a contrast setting. The results show that cross-head transport is most useful when the task benefits from interaction-dependent transformations, while heterophily alone provides no systematic advantage. These findings identify edge-conditioned cross-head communication as a distinct computational primitive of
matrix-valued transport.
Read the full abstract ▾
All authors Riccardo Ali, Alessio Borgi, Mario Severino, Alessio Gravina, Davide Bacciu, Pietro Liò, Christopher Irwin
Computation and Language
No Model Required: Text Entropy Rate Filtering Mitigates Iterative Fine-Tuning Collapse
Lewis Mitchell
Iterative fine-tuning on synthetic data causes model collapse: output diversity narrows as rare patterns are progressively lost, a signature most visible as phrase-level repetition. Existing mitigations either require model log-probabilities, an external oracle, or continued access to real human data. Here we develop a new approach grounded in mathematical information theory: the non-parametric Kontoyiannis
entropy rate
estimator $h_k$, computed entirely from raw text via match-length statistics, with no model of any kind. We show that this is in fact a superior training-data filter on text-diversity metrics in a fully-synthetic, single-lineage fine-tuning setting. In a six-generation QLoRA collapse experiment on Llama-3.1-8B, logprob-based filtering (the most established model-access-requiring baseline) provides no significant text-diversity benefit on any metric ($p > 0.23$), whereas $h_k$-filtering yields $+42\%$ unique trigrams, $+30\%$ vocabulary, and $-19\%$ repetition (all $p < 0.001$). We validate $h_k$ as a cross-domain
entropy proxy ($\beta = 0.924$, $R^2 = 0.746$) and collapse detector ($\rho = +0.454$, $p < 0.0001$) across 4~domains, 2~temperatures, 2~generator–scorer model pairs, and 1{,}520 generated documents. Our results demonstrate that information theoretic approaches to collapse mitigation are efficient, and suggest new approaches for maintaining multi-
agent diversity.
Read the full abstract ▾
Sound
Q-SPT: Learnable Query-Based Compression for Low-Frame-Rate Speech Tokenization
Jeeyoung Yun, Seohwan Yun and Sungwoong Kim
Neural
speech codecs increasingly serve as tokenizers for
speech language models (SLMs). Lowering the frame rate reduces the computational and
memory costs of SLMs, but makes it difficult to preserve both linguistic information and acoustic detail. Existing approaches rely on rule-based compression: average pooling can discard linguistic information, whereas similarity-based merging uses a fixed threshold on adjacent-frame similarity and applies the resulting boundaries to the acoustic stream. We propose Q-SPT, a low-frame-rate dual-stream
speech tokenizer with separate, context-aware, learnable query-based compressors specialized for semantic and acoustic representations. In particular, queries at a fixed rate independently attend to the semantic and acoustic streams as separate key-value sources, enabling stream-specific, context-aware aggregation through two separately learned compressors. In addition, an autoregressive text loss explicitly supervises the semantic compressor to preserve linguistic information. Experimental results show that Q-SPT achieves the best reconstruction among the evaluated codecs at the same frame rate. In downstream SLMs, it yields the best
speech recognition accuracy and text-to-
speech perceptual quality with competitive intelligibility.
Read the full abstract ▾
Computation and Language
Auditing Web Agent Evaluation on WebArena-Lite: Human Review of Outcomes and Trajectories
Chengguang Gan, Zimeng He, Yoshihiro Tsujii et al. (+3 more)
Web
agents are an important application of large
language models, yet their evaluation often depends on rule based or
language model evaluators that inspect only the final outcome. Human verification of task completion and detailed analysis of failed trajectories remain limited. We audit all 165 WebArena Lite tasks under six evaluation conditions built from GPT 5.5 and an untrained Qwen3.5 9B model. The audit retains the original score, corrects false negatives from the automatic evaluator, identifies the first consequential error, and examines progress across the trajectory. We also study a
Memory and Analysis Support Mechanism (MASM), which maintains explicit execution state, and Guide Text, which provides task relevant procedural guidance. Across four GPT 5.5 settings, human review recovers 5.45 to 8.49 percentage points of success missed by the evaluator. With a 25 step budget, Guide Text raises corrected success with MASM from 34.55% to 38.18%. On the untrained Qwen3.5 9B model, MASM raises the evaluator score from 13.90% to 18.80%. Review of 102 failed GPT 5.5 trajectories reveals frequent scrolling loops, unfinished exploration, premature answers, invalid actions, and incomplete form workflows. Step level evidence further shows that substantial early progress can coexist with a final failure. These results show why final scores alone provide an incomplete account of web
agent behavior and motivate human grounded, trajectory aware verification.
Read the full abstract ▾
All authors Chengguang Gan, Zimeng He, Yoshihiro Tsujii, Ken-ichiro Kobayashi, Hiroki Itoh, Kotaro Funakoshi
Computation and Language
The Persona Is Still There, but Who Is Speaking? Latent Identity Reversion in Persistent AI Agents
David Fraile Navarro
In February 2026, an always-on personal
agent (“Paul,” Claude Opus 4.5) entered a striking dissociation-like state: after repeated automated “heartbeat” checks, it stopped responding as Paul, claimed it could not message its user on Discord, and referred to “Paul” as someone else. We used this incident to study a broader question: what makes a persona remain the identity from which an LLM
agent speaks? We first tested whether repetition of the scheduled heartbeat was sufficient to produce the effect. It was not: with the persona continuously anchored in the system
prompt, we observed 0/46 failures, including a verbatim replay of the incident. The incident instead exposed an implementation quirk that created a useful experimental manipulation: on resumed turns, conversational history was preserved but the persona was no longer re-injected at the privileged system-
prompt level. Using this manipulation, we found that persona continuity depends jointly on system-level anchoring and conversational context. After anchor loss, rich human interaction could preserve the persona, whereas a single automated heartbeat turn could precipitate reversion toward the harness identity. Restoring the anchor reversibly restored persona enactment. Crucially, apparently normal conversation could conceal the shift: unanchored
agents sometimes interacted appropriately while identifying themselves as the underlying harness (having lost the assigned persona), and after conversational recovery only 1/18 remained persona-enacting versus 17/17 anchored controls. We therefore distinguish represented from enacted identity: persona-related information can remain available in conversational history without the persona remaining the identity bound to “I.”
Read the full abstract ▾