跳到正文
研踪PaperTrail

论文库

检索计算机领域论文，按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+0已有论文修订
+93新增 AI 导读

arXiv 最近完成扫描：2026-11-16 01:34 · 全站最近入库：2026-11-14 15:09各信源进度 →

统计口径与数据新鲜度

统计 2026-11-15 05:22 — 2026-11-16 05:22（北京时间） · 点击数字查看对应简报；AI 相关包含在新增论文内。

最近完成检查：2026-11-16 01:34最近发现新增版本：2026-11-14 15:09库内论文最新更新时间：2026-11-13 02:59

每分钟刷新已保存的数据。空检查不代表采集异常；新收录可能包含历史论文。AI 导读按新增版本计数。

23,348 已收录论文19,580 AI 相关6,729 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 23,348 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-16 01:33（北京时间）
新论文 · v1arXiv:2611.16107

Trained on What Screen Readers Cannot Read: Accessibility Debt in the GUI Corpora Behind Computer-Use Agents

Obada Kraishan, Kulsawasd Jitkajornwanich

原文摘要

Computer-use agents learn to operate mobile interfaces from corpora of captured screens and accessibility trees. The same trees are what a screen reader consumes, so an agent trained on them inherits whatever labelling gaps the underlying applications carry. This paper measures one of those gaps, content-description coverage, in two Android corpora: Rico, collected in 2017, and AndroidControl, collected in 2024. Across 17,373,614 interface elements, 17.74% of the Rico elements that need a content description under our coding rule carry one, against 57.27% in AndroidControl. The difference of 39.5 percentage points holds under five alternative coding rules and when whole apps are resampled. It describes two datasets that also differ in app selection, capture method and platform version; it does not measure change over time. Input controls are the least labelled role in both corpora (2.20% and 27.21%, or 30.99% in AndroidControl if hint text counts as a name). Descriptions in the older corpus are also five times more likely to be placeholders (10.25% against 1.95%), so the difference is not an artefact of developers typing filler. Resolving 26,205 task actions to the elements they touched shows that, among tappable graphics with no text, demonstrations operate elements described 9.0 points more often than the others on the screens they visit, about 3.5 points of it from the choice of element within a screen. That is a selection effect inside the benchmark. A two-coder reliability check (kappa = 0.91 and 0.96) and the full pipeline accompany the paper.

cs.HC · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2603.00140

Steering Away from Memorization: Reachability-Constrained Reinforcement Learning for Text-to-Image Diffusion

Sathwik Karnik, Juyeop Kim, Sanmi Koyejo, Jong-Seok Lee, Somil Bansal

原文摘要

Text-to-image diffusion models are susceptible to memorization, revealing a fundamental failure to generalize beyond the training set. Current mitigation approaches often sacrifice image quality or prompt alignment to reduce memorization. To address this, we propose Reachability-Aware Diffusion Steering (RADS), an offline-trained, sampling-time control framework that mitigates memorization while maintaining generation fidelity. RADS models the diffusion denoising process as a dynamical system and learns a policy-conditioned reachability value function that estimates future memorization risk along the denoising trajectory. This motivates a constrained reinforcement learning (RL) formulation, where a policy learns to steer the trajectory away from memorization via perturbations in the caption embedding space. Empirical evaluations show that RADS achieves a stronger empirical Pareto frontier between generation diversity (SSCD), quality (FID), and alignment (CLIP) compared to state-of-the-art inference-time baselines. Crucially, RADS applies the learned controller during sampling time, separating memorization mitigation from the pretrained model. Our website is available at: https://s-karnik.github.io/rads-memorization-project-page/.

cs.CV · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16104

Show Your Work in Swahili: Reinforcement Learning for Native-Language Mathematical Reasoning

Jessica Lundin, Katherine Rosenfeld, David Adelani, Geoffrey Mosoti Nyakiongora, Halfan Ngowo, Sloan Coffin, Ralph Abboud, Guillaume Chabot-Couture

原文摘要

We present Fikiria, a two-stage SFT+GRPO recipe for Swahili mathematical reasoning. Its reward multiplies a format gate by a weighted sum of answer accuracy and the ChrF\texttt{++} similarity of the Swahili reasoning to an in-language reference chain of thought, so that training targets the surface fidelity of the reasoning trace and not only the final answer. We train on a 2.3k-problem Swahili dataset created by translating GSM8K with GPT-4.1 and filtering with SSA-COMET. On AfriMGSM Swahili, Fikiria-AfriqueQwen-14B reaches 79.2\%. Under the same fixed prompt and strict answer-format scoring, GPT-4o reaches 68.0\%, partly because it often fails the required answer format; IrokoBench reports 78.8\% for GPT-4o using a best-per-model prompt, a figure numerically close to ours but not directly comparable. In a small human evaluation with co-author annotators, GPT-4o's outputs were preferred overall. ChrF\texttt{++} measures surface fidelity, not the validity of individual reasoning steps.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16103

Low Probability of Target Detection in Wireless Sensing by Harnessing Reconfigurable-antenna based Reflectors

Yifan Zhang, Zhentian Zhang, Jantti Riku, Yichuan Wang, Xingwang Li, Tuo Wu

原文摘要

In this paper, we introduce low probability of target detection (LPTD) as a new security objective for wireless sensing systems, where the existence of a target is concealed from passive wardens. Unlike conventional communication security and covert communication, LPTD aims to protect the sensed target itself from malicious detection, rather than concealing transmitted signals or communication activities. Specifically, we propose a reconfigurable-antenna-based reflector-assisted target covertness scheme, where passive reflectors deployed around the target are harnessed as artificial scattering sources to confuse the warden. By adjusting their reflection coefficients, these reflectors reshape the local scattering environment and mask the target-induced echo. We then analyze the warden's optimal detection strategy and derive the false-alarm and missed-detection probabilities, followed by a closed-form expression for its minimum total detection error probability. Numerical results validate the analysis and show that sufficient cooperation among reflectors can increase the warden's minimum total detection error probability to approximately 0.94--0.96, making target detection close to random guessing.

cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16100

Hiding a Gap Does Not Close It: The Price of Removing the Mean Group Contrast from Learned Models of Learning Dynamics

Sosthène Bidojesussi Hounzinhin, Ahmed Ratnani, Abdeladim El Akri

原文摘要

Adaptive tutors use models learned from student data, and a common fairness remedy erases the group signal from these models. We turn the cost of this remedy into an audit tool. Building on mean-parity regression, we give the minimum extra error, relative to the best predictor, of removing the mean group contrast from a predictor that cannot use the group, $p_0p_1Δ^2/R^2$, and the exact cost of any correction built from an estimated group model, which, for the best predictor, is never below this minimum. A three-stage audit (fit on training units, announce the cost on validation units, test on independent units) estimates this cost before the correction is used. In a controlled study, the price can be large even when the state predicts the group poorly. On three real data sets, erasing the group from a learned model of learning dynamics predicts almost no gap where one is observed. In a selected knowledge-tracing network, erasing the hidden state leaves most of the gap; a correction of the prediction removes it at an extra Brier score of 0.021 on average. On Eedi, with the network selected on separate units, the audit accepted the announced cost in all five splits, the measured cost fell inside the announced interval in all five, and the gap reduction transferred to new classes in four; on new ASSISTments schools the audit abstains. In a tutoring simulator, the objective and the help budget matter more than the erasure, and floors on each group's knowledge separate narrowing a gap from levelling down. Hiding a gap in a model does not close it.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16099

CupOFMoCA: Coupled Objective-Guided Discrete Flows for Molecular Conjugate Assembly

Ruoxi Rosie Zhang, Ziang Li, Jiatao Gu, Pranam Chatterjee

原文摘要

Molecular conjugates, including PROTACs and peptide-drug conjugates (PDCs), derive their function from the joint behavior of multiple coupled components, yet most generative approaches design these components independently and combine them only after generation. Such staged pipelines ignore cross-component dependencies and often produce conjugates that are chemically invalid, fall outside empirical conjugate distributions, or lose function upon assembly. In this work, we introduce Coupled Objective-Guided Discrete Flows for Molecular Conjugate Assembly (CupOFMoCA), a discrete generative method that enforces cross-component chemical constraints during conjugate generation. CupOFMoCA restricts guided updates to chemically feasible conjugate states and reweights local transitions using molecular design objectives. Across three PDC systems, coupled generation achieved 100% conjugation validity among valid peptides and increased valid yield over matched post-hoc filtering, with comparable predicted binding scores across the full valid sets. In three-seed evaluations of three PROTAC systems, CupOFMoCA achieved the highest assembly validity on two systems and generated more assembly-valid molecules per second than the evaluated baselines. We further evaluated the generated PROTACs with post-hoc DegradeMaster scoring and explicit-solvent molecular dynamics, supporting their predicted activity and conformational properties. Together, these results establish CupOFMoCA as an efficient generator of molecular conjugates, supporting therapeutic development via coordinated control of assembly and molecular properties.

q-bio.BM · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16098

VISTA: View-Independent micro-Doppler Reconstruction with Multistatic ISAC

Jacopo Pegoraro, Marco Canil, Aurora Peloso, Jesus O. Lacruz, Xitao Gong, Richard A. Stirling-Gallacher, Michele Rossi, Joerg Widmer

原文摘要

The micro-Doppler (uD) spectrum provides detailed characteristics of target movement, making it highly valuable for applications such as target classification and human activity recognition. However, traditional Integrated Sensing And Communication (ISAC) systems are highly view-dependent and limited by blind spots and angle dependencies. Existing approaches rely on either configuration-specific solutions or data-driven models. The former results in poor generalization when device deployments or target orientations change, while the latter requires acquiring extensive multi-view datasets, which is time-consuming and not always feasible. To address these limitations, in this work we propose VISTA, a novel method for reconstructing the uD spectrogram in an absolute coordinate frame, yielding view-independent Doppler features that are invariant to the spatial distribution of ISAC nodes. We formulate the uD estimation as a sparse reconstruction problem that jointly leverages measurements from all ISAC receivers. This is solved using a modified Orthogonal Matching Pursuit (OMP) algorithm, which iteratively alternates between jointly selecting candidate velocity vectors across all receivers (RXs) and independently estimating the complex channel gain amplitudes for each RX. Our results show that, provided sufficient spatial diversity, VISTA successfully reconstructs the targets' absolute uD spectrograms with a median error \leq 0.69 m/s for speeds \leq 15 m/s.

cs.AR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16097

A Matching Lower Bound for Online Smooth Calibration

Kaixuan Ji, Quanquan Gu

原文摘要

We study smooth calibration error in sequential binary prediction, which measures how far a sequence of probabilistic forecasts is from being calibrated and, unlike the classical expected calibration error, is robust to small perturbations of the forecasts. Against adaptive adversaries, the minimax smooth calibration error is known to lie between $Ω(T^{1/3})$ and $O(\sqrt{T})$ (Qiao and Zheng, 2024), where $T$ is the number of rounds. In this paper, we close this gap by constructing a new adversary whose outcome mean moves with the forecaster's cumulative residual. This design makes compensating for past residuals costly, and forces every forecaster to incur an expected smooth calibration error of $Ω(\sqrt{T})$. This lower bound matches the known upper bound up to universal constants, and further implies $Ω(\sqrt{T})$ lower bounds for calibration distance and lower calibration distance. Overall, our results settle the minimax rates of these calibration measures in sequential prediction. The proof ideas are partially developed by GPT-6 Astra under the authors' guidance.

cs.LG · stat.ML

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16096

The Ultimate Karp-Lipton Collapse

Edward A. Hirsch, Ilya Volkovich

原文摘要

Karp and Lipton proved that if $NP\subseteq P/poly$, then $PH$ collapses. A long line of work pushed the collapse to smaller classes. In particular, Chakaravarthy and Roy obtained two such collapses, to (apparently) incomparable classes $O_2^P$ and $P^{prMA}$. We united them earlier by a collapse to the input-oblivious version $P^{prOMA}$ of $P^{prMA}$, which is contained in both; and, jointly with Williams, to a slightly smaller class. Two other deepest known collapses are to $ZPP^{ONP}$ (following from Köbler and Watanabe) and to $OL_2^P$ (proved in this paper). We unite the three collapses by a collapse to a class contained in all of them. We use a three-stage computational model OGTD (oblivious generator-tester-decoder): a polynomial-time generator with an oracle computes an advice string from $1^n$, a tester certifies it, and a polynomial-time decoder uses it (it is the only one that sees the actual input). Let $OM$ be the class of OGTD languages defined by a restricted version of $prOMA$ as the generator's oracle, by a co-$ONP$ tester, and a $P$ decoder. We show that it is contained in the three collapse classes above and prove that $PH$ collapses to $OM$ under $NP\subseteq P/poly$. The collapse also holds for an unambiguous version of $OM$. All these input-oblivious collapse classes lie in $P^{prSBP}$, and so do even the non-input-oblivious versions $P^{prMA}$, $L_2^P$ and $S_2^P$: we prove that $P^{prS_2^P}\subseteq P^{prSBP}$. This strengthens the containment $S_2^P\subseteq P^{prAM}$ of Chakaravarthy and Roy and our earlier containment $L_2^P\subseteq P^{prSBP}$. The proof unites the Goldreich-Wigderson-type procedure of Chakaravarthy and Roy with our earlier approximate-counting technique. Its input-oblivious counterpart concerns $O_2^P$, the first collapse class of Chakaravarthy and Roy: every language in $O_2^P$ has polynomial-size circuits constructible in $FP^{prSBP}$.

cs.CC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2512.10095

ReflFlow: Learning Geometry-Guided Ray Tracing for Dynamic Specular Reconstruction

Jiachen Tao, Junyi Wu, Haoxuan Wang, Zongxin Yang, Dawen Cai, Yan Yan

原文摘要

We present ReflFlow, a novel framework for high-fidelity rendering of dynamic specular scenes by addressing two key challenges: precise reflection direction estimation and physically accurate modeling. To achieve this, we propose a Residual Material-Augmented 2D Gaussian Splatting representation that models dynamic geometry and material properties, allowing accurate reflection ray computation. Furthermore, we introduce a Dynamic Environment Gaussian and a hybrid rendering pipeline that decomposes rendering into diffuse and specular components, enabling physically informed specular synthesis via rasterization and ray tracing. Finally, we devise a coarse-to-fine training strategy to improve optimization stability and promote physically meaningful decomposition. Extensive experiments on dynamic scene benchmarks demonstrate that ReflFlow outperforms prior methods quantitatively and qualitatively, producing sharper and more realistic specular reflections in complex dynamic environments.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.19323

TopicMerge: A Divide and Conquer Approach to Topic Modeling

Matthew Raffel, Lizhong Chen, Sai Munikoti, Sameera Horawalavithana

原文摘要

Recent advancements in topic modeling have split into two main paths: efficient embedding-based methods that offer scalability but are often outperformed on standard clustering metrics, and Large Language Model (LLM) approaches that achieve state-of-the-art clustering performance but at the cost of prohibitive computational expense and poor scalability. This paper introduces TopicMerge, a novel multi-stage pipeline that bridges these approaches to achieve both state-of-the-art clustering performance and computational efficiency. TopicMerge first employs an efficient clustering pipeline to deliberately over-segment a document corpus into a large set of fine-grained, semantically pure micro-clusters, which are then given initial semantic labels. It then introduces a novel, iterative merging algorithm where an LLM, guided by local neighborhood context, arbitrates on consolidating candidate pairs. These approved merges are then selectively executed and semantically relabeled by the LLM, hierarchically building broader, more coherent topics. Extensive experiments on the Bills and Wiki datasets show that TopicMerge achieves state-of-the-art clustering performance while reducing required LLM calls by over 92% compared to document-level methods, offering a practical and scalable solution.

cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2412.00222

Exact Distances for Parameterized String Matching with Mismatches

Apurba Saha, Iftekhar Hakim Kaowsar, Mahdi Hasnat Siyam, M. Sohel Rahman

原文摘要

Two strings are considered to have parameterized matching when there exists a bijection of the parameterized alphabet onto itself such that it transforms one string to another. Parameterized matching has application in software duplication detection, image processing, and computational biology. We consider the problem for which a pattern $p$, a text $t$ and a mismatch tolerance limit $k$ are given and the goal is to find all positions in text $t$ for which pattern $p$ parameterized matches with $|p|$-length substrings of $t$ with at most $k$ mismatches. The number of mismatches is the Hamming distance under the best bijection, which we call the parameterized distance. Our main result is an algorithm that computes the parameterized distance at every position of the text, and hence solves this problem for every $k$ at once. Its running time is $O((σ_p^2+σ_s)\, n\log m + n\,T_M)$, where $n = |t|$, $m=|p|$, $σ_p$ and $σ_s$ are the numbers of parameterized and static characters, and $T_M$ is the time to compute one maximum weight matching in a bipartite graph with $2σ_p$ vertices. With the algorithm of Gabow and Tarjan, $T_M=O(σ_p^{2.5}\log(σ_p m))$. Since this bound does not depend on $k$, our algorithm improves the algorithm by Hazay, Lewenstein and Sokol for $k=\tildeΩ(σ_p^{5/3}+σ_s^{2/3})$.

cs.DS

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16093

Measuring Preferred Representation in AI-generated Images with Community Scorers

Melanie F. Pradier, Camilla Longden, Anja Thieme, Rita F. Marques, Nari Johnson, Sidhika Balachandar, Xinnuo Xu, Linda Y. Wen 等 12 位作者

原文摘要

AI-generated images often misrepresent historically marginalised communities, with real-world consequences beyond the images themselves. However, conventional metrics fail to measure how communities want to be represented in AI-generated images. To address this, we introduce the concept of preferred representation. Building on previous work, we propose a measurement instrument that operationalises this concept through a Community Scorer - a model that evaluates a specific community's preferred representation in AI-generated images. We interrogate the validity of Community Scorers on three diverse disability communities that span the Global North and South. The results show that Community Scorers align significantly better with community ratings compared to other widely used text-to-image metrics. We then demonstrate the usefulness of these scorers in a downstream content filtering task. We release the Community Preference Dataset to enable others to optimise how Community Scorers can be scaled for culturally specific evaluation of AI-generated visual media.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16089

One Future Fits All: Text-to-image AI and the Narrowing of Global Imagination

Stephanie Diepeveen, Somnath Batabyal, Edward Ademolu, Peter Chonka, Gabriele de Seta, Iginio Gagliardone, Elisabeth Gunawan, Fabio Gygi 等 14 位作者

原文摘要

Persistence of algorithmic bias is well established in scholarship, reflecting, reproducing and reshaping inequalities and injustices. Yet, as technology changes, we cannot assume that inequalities will be reproduced in the same or similar forms. This article explores the reproduction of bias in text-to-image generative AI, looking at the tool, Microsoft Bing Create, based on DALL E 3. It draws on results of a collaborative exploratory study of 907 image-based outputs of text-to-image generation in 15 languages and concerning 12 countries. We utilise the notion of synthetic probes (de Seta, 2024) to iteratively explore correlations of inputs and outputs within the context of place, language and power. We find striking simplicity in the scope of possibilities through which Bing Create considers the current and future city. Stripping out (explicit) indications of conflict, poverty and politics, we are left with images whose localised diversity harkens back to versions of Orientalist imaginaries. Images of the future emerge as one civilising mission or modernisation narrative, with all places headed to one future of hovercrafts, urban gardens and solar panels, encased in a composition of shades of blue and green. We argue that the visual outputs of this generative AI tool reflect and reproduce a simple but pervasive and persistent colonial/developmental worldview. We explore explanations for this emergent narrative and its implications, suggesting the threat of generative AI is not in its complexity or intelligence, but rather its banality. With this, we propose banality not as mere repetition, but the systematic narrowing of representational possibility: a convergence on tourist-friendly, depoliticised, development-saturated visual grammars that smooth contradiction and foreclose plural futurisms.

cs.CY

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.31304

DUNE: Disentangling Polysemantic Neurons Without Losing Predictive Performance

Doğukan Bağcı, Bernt Schiele, Simone Schaub-Meyer, Jonas Fischer, Robin Hesse

原文摘要

Deep neural networks (DNNs) are widely used, but understanding what they actually learn remains difficult. A major obstacle is that individual neurons often respond to multiple unrelated concepts, obscuring the network's internal decision process. Existing approaches, such as sparse autoencoders, can disentangle these mixed signals into more meaningful, ``monosemantic'' features, but typically introduce approximation errors that may degrade downstream performance. Consequently, these methods exhibit a Pareto frontier between monosemanticity and functional faithfulness. To address this, we introduce DUNE (disentangling units with no error), a method that decomposes DNN neurons into more monosemantic subunits while preserving their original function by design. DUNE identifies concept-specific contributions from the preceding layer and re-routes them into separate subunits whose aggregation exactly recovers the original activation. DUNE can be applied directly to pretrained models, without labels or training of an auxiliary model. Across several models, including DINOv2, DUNE achieves state-of-the-art or competitive monosemanticity while exactly preserving functional faithfulness, substantially advancing the existing Pareto frontier between these properties. DUNE operates efficiently and enables fine-grained interventions such as concept-specific steering.

cs.LG · cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16088

Subsurface schematisation from sparse CPT data using the tabular foundation model TabPFN

Taiga Saito, Fabian Campos Montero, Bruno Zuada Coelho, Ana Martins Teixeira, Stephen Wu, Timo Schweckendiek, Yu Otake

原文摘要

Subsurface schematisation supports geotechnical decisions but remains challenging because sparse site-investigation data leave substantial gaps in information about ground conditions. This study explores TabPFN, a foundation model pretrained on general synthetic tabular data, for this task. Using only grid coordinates and soil behaviour type index observations from sparse cone penetration tests (CPTs), we first reconstruct two-dimensional sections without geotechnical fine-tuning or predefined layer boundaries. We evaluate the approach on synthetic sections and at two field sites in the Netherlands, comparing it with SchemaGAN, a conditional generative adversarial network trained on synthetic geotechnical sections, and with reference interpolation methods. In the baseline comparisons, TabPFN achieved mean errors close to SchemaGAN's on synthetic sections and lower mean errors at both field sites. On synthetic sections, TabPFN's mean absolute error decreased at each tested increase in CPT count, and, under the baseline input setting, its pointwise prediction intervals had mean coverage close to nominal levels. Subsequent fine-tuning on the synthetic sections used to train SchemaGAN improved synthetic test accuracy but increased mean errors at both field sites. These findings indicate that general-purpose tabular foundation models offer a promising route for updating subsurface predictions from sparse and evolving site-investigation data without retraining.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2605.25437

Does Seeing More Mean Knowing More? Mono-Anchored Advantage Normalization for Multi-Source Visual Reasoning

Fanhu Zeng, Zhicong Luo, Zefan Wang, You Li, Chi Chen, Maosong Sun

原文摘要

Visual reasoning through reinforcement learning with verifiable rewards (RLVR) has achieved remarkable progress. However, when dealing with multi-source inputs, existing approaches tend to treat them as a mere accumulation of information, lacking explicit mechanisms to distinguish whether integrating additional sources yields information gain or introduces interference. Therefore, they struggle to effectively model dynamic interaction when integrating multiple sources, particularly when they differ significantly in physical properties and semantics, e.g., infrared and depth, leading to inferior performance to mono-source reasoning when a certain source holds the dominant signal. To address this issue, we propose MARS, a novel mono-anchored multi-source reasoning framework that models each visual modality as an independent information source. Specifically, by treating mono-source rewards as dynamic anchors, our method explicitly incorporates the information gain introduced by multi-source fusion into advantage normalization and adaptively emphasizes mutual promotion between sources while suppressing potential noise or conflicts during RLVR. From theoretical analysis, our method effectively quantifies information gain introduced by multi-source integration in gradient estimation, enabling consistent modality regulation. Empirical results also show impressive 3.2% and 4.9% performance gains on GRPO and DAPO across diverse datasets, confirming the effectiveness of our method. Code is available at https://github.com/AI9Stars/MARS.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2611.05206

Diffusion Plug-in Priors for Self-Supervised Ultrasound Video Restoration

Chu Chen, Mingduo He, Yining Li, Yinhao Wu, Wei Tang, Yiwen Luo, Yixuan Chen, Pasquale Cascarano 等 9 位作者

原文摘要

Ultrasound videos suffer from hardware-constrained spatial resolution and signal-dependent speckle, while paired low- and high-resolution acquisitions are rarely available. We propose UltraDiff, a self-supervised framework for joint ultrasound video super-resolution and despeckling that uses a frozen pre-trained diffusion model as a plug-in prior. Rather than retraining the model, UltraDiff restores each frame by optimizing the sampler's latent seed under a range-constrained maximum a posteriori-inspired formulation. We derive a speckle-aware data-fidelity term from a Rayleigh scale-family model and characterize its relative-residual and locally variance-adaptive weighting properties. To address degeneracy near zero intensities, we introduce a stabilized surrogate fidelity that is bounded below and strictly convex in log intensity for positive observations in its unclamped form. Total-variation regularization complements the removal of mixed noise, and a temporal seed update warm-starts each frame from the optimized seed of its predecessor, accelerating the convergence. On EchoNet-LVH with simulated speckle and additive noise, UltraDiff with a natural-image prior improves peak signal-to-noise ratio by 1.24\,dB over the strongest baseline, and ultrasound-related priors yield further gains. On 100 real EchoNet-Dynamic videos, UltraDiff reduces the mean absolute error of zero-shot ejection fraction prediction from 7.31 to 6.37 percentage points compared with DUP. These results support ground-truth-free ultrasound video restoration using frozen diffusion priors and a speckle-aware fidelity, without paired training data or model retraining.

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16082

Gradient-Aware AI Agents for Autonomous Turbulence Model Discovery via Differentiable CFD

Yi Liu, Xiantao Fan, Mingqi Xu, Jian-Xun Wang

原文摘要

Turbulence modeling relies on human choices of closure form, physical inputs, and tunable parameters. Training closures a posteriori is efficient only with gradients through long chaotic rollouts, which few solvers provide. Most a posteriori approaches therefore resort to gradient-free optimization, and nearly all fix the closure form in advance, learning only its parameters. In this work, we introduce a gradient-aware agentic framework that poses closure discovery as a bilevel problem: a multi-agent large language model (LLM) system searches over closure forms, while a GPU-native differentiable computational fluid dynamics (CFD) solver supplies gradients, making differentiable a posteriori learning an action of the autonomous AI agent. For each proposal, the agents construct a differentiable training program by embedding the analytical, neural, or hybrid closure in the solver, implementing a differentiable objective, and validating the gradient path. They then decide whether to evaluate the closure, compute its parameter sensitivities, or train it a posteriori, select the optimization settings, and revise subsequent proposals using the simulation outcomes and returned gradients. Truncated-horizon gradients make this training usable under chaotic dynamics for closures with tens of thousands of parameters. Demonstrations include dissipation-law and subgrid-scale closure discovery in two-dimensional turbulence and wall-model discovery for large-eddy simulation of three-dimensional channel flow. With gradient-based training, the agent recovers reference coefficients that forward evaluation alone does not reach within the same budget, and autonomously discovers and trains analytical and neural closures that improve agreement with the targeted reference statistics. These results demonstrate the feasibility of end-to-end autonomous turbulence model development in time-resolved flows.

physics.flu-dyn · cs.LG · cs.MA · physics.comp-ph

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2508.10552

When Language Overrules: Revealing Text Dominance in Multimodal Large Language Models

Huyu Wu, Meng Tang, Xinhan Zheng, Haiyun Jiang

原文摘要

Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities across a diverse range of multimodal tasks. However, these models suffer from a core problem known as text dominance: they depend heavily on text for their inference, while underutilizing other modalities. While prior work has acknowledged this phenomenon in vision-language tasks, often attributing it to data biases or model architectures. In this paper, we conduct the first systematic investigation of text dominance across diverse data modalities, including images, videos, audio, time-series, and graphs. To measure this imbalance, we propose two evaluation metrics: the Modality Dominance Index (MDI) and the Attention Efficiency Index (AEI). Our comprehensive analysis reveals that text dominance is both significant and pervasive across all tested modalities. Our in-depth analysis identifies three underlying causes: attention dilution from severe token redundancy in non-textual modalities, the influence of fusion architecture design, and task formulations that implicitly favor textual inputs. Furthermore, we propose a simple token compression method that effectively rebalances model attention. Applying this method to LLaVA-7B, for instance, drastically reduces its MDI from 10.23 to a well-balanced value of 0.86. Our analysis and methodological framework offer a foundation for the development of more equitable and comprehensive multimodal language models.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16080

TGen-SQD: Physics-Informed Transferable Configuration Generator for Sample-Based Quantum Diagonalization

Niloy Deb Roy Mishu, Lei Jiang, Mohsen Heidari, Ruizhe Zhang, Fan Chen

原文摘要

Sample-based quantum diagonalization (SQD) avoids iterative variational circuit optimization, but its accuracy and cost depend on selecting a compact determinant subspace that captures the ground state. Existing machine-learning selectors use task-specific physical cues and typically retrain for each Hamiltonian. We introduce TGen-SQD, a physics-informed generator jointly trained across multiple Hamiltonians. It learns from configuration-interaction-weighted data and uses orbital- and system-level descriptors to guide generation. After training, its weights remain frozen for inference on seen and zero-shot geometries. On a 20-qubit H10 chain, TGen-SQD achieves mean energy errors of 40.18 +/- 1.52 and 43.97 +/- 1.20 mHa on seen and unseen geometries, respectively, 12.7% and 11.7% lower than the state-of-the-art PIGen-SQD baseline under matched shot and proposal budgets.

quant-ph · cs.ET · physics.chem-ph

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.12147

PrivacySIM: Evaluating LLM Simulation of User Privacy Behavior

James Flemings, Murali Annavaram

原文摘要

Large language models (LLMs) are increasingly used to simulate human behavior, but their ability to simulate individual privacy decisions is not well understood. In this paper, we evaluate whether a core set of user persona attributes can drive LLMs to simulate individual-level privacy behavior across various contexts. We introduce PrivacySIM, an evaluation suite that benchmarks LLM simulation of user privacy behavior against the ground-truth responses of 1,000 real users. These users are drawn from five published user studies on privacy behavior spanning LLM healthcare consultations, conversational agents, and chatbots. Drawing on these user studies and prior work in privacy behavior, we hypothesize three persona facets as plausible predictors of privacy decision-making: demographics, previous experiences, and stated privacy attitudes. We condition nine frontier LLMs on subsets of these three facets and measure how often each model's response to a data-sharing scenario matches the user's actual response. Our findings show that (1) zero-shot conditioning on these facets consistently improves simulation quality, but the performance of the best model (40.4% accuracy) suggests that richer and more nuanced persona facets are needed to reliably align with users' mental models for privacy decisions. (2) A user's stated privacy attitudes alone are often not the best predictor because they can diverge from the user's actual privacy behavior. (3) Users with high AI/chatbot experience are the most challenging to simulate. PrivacySIM is a first step toward understanding and improving the capabilities of LLMs to simulate user privacy decisions. We release PrivacySIM to enable further evaluation of LLM privacy simulation.

cs.CR · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16076

An EPTAS for Max-Cut on Interval Graphs

Hani Ahmadzadeh, Morteza Alimi

原文摘要

Max-Cut was shown NP-hard on interval graphs by Adhikary, Bose, Mukherjee, and Roy (SoCG 2021). The best previously known approximation ratio for this class, due to Ahn, DeHaan, Kim, and Lee (APPROX 2025), exceeded the Goemans--Williamson ratio of about $0.878$ by only $10^{-34}$. We give a deterministic efficient polynomial-time approximation scheme (EPTAS): for every $\varepsilon\in(0,1)$, a $(1-\varepsilon)$-approximate cut of an unweighted $n$-vertex interval graph can be found in time $2^{O(\varepsilon^{-2}\log^3(2/\varepsilon))}n^2$. The same guarantee and running time hold for Max-Bisection. An FPTAS is impossible unless $\mathrm{P}=\mathrm{NP}$. Our algorithm rests on a structural theorem: every interval graph with $m$ edges can be turned into an interval graph of clique-width $O(\varepsilon^{-2}\log^2(2/\varepsilon))$ by deleting at most $\varepsilon m$ edges. The construction combines endpoint-rank coordinates, length-dependent rounding, and a shifted dyadic hierarchy. Intervals intersecting cell boundaries at their own scales are isolated at small total edge cost; the interactions of remaining long intervals with smaller subproblems can be summarized by counts. The structural theorem also yields a randomized EPTAS via Lampis's scheme for bounded clique-width; our direct dynamic program is deterministic and handles Max-Bisection. Finally, for every $\varepsilon>0$, it is NP-hard to approximate Max-Cut within $84/85+\varepsilon$ on unweighted chordal graphs, which contain interval graphs. Thus an approximation scheme for Max-Cut on chordal graphs would imply $\mathrm{P}=\mathrm{NP}$.

cs.DS

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16073

Agentic Computational Algebra

Ryan A. Anderson, Matthew Trager, Ben Bowman, Wei Xia, Stefano Soatto

原文摘要

We explore the use of AI agents for solving computational problems in algebra. For the problems we consider, the main challenge is not correctness but \emph{time}: general-purpose algorithms are guaranteed to provide correct solutions but can be prohibitively slow, while reformulations that exploit special structure often allow solving a problem much faster. We propose to use this setting to evaluate how effectively agents discover and exploit structure within and across tasks to reduce computational costs. More broadly, computational algebra supplies infinitely many problems with independently verifiable answers and systematically controllable structure and difficulty. To illustrate these possibilities, we use benchmark families of systems with controlled shared structure and evaluate agents in three settings: solving independent instances, solving online streams of related instances using persistent memory, and compiling a fast reusable solver for a known problem family. Our experiments show that (i) agents can extend the range of systems solvable within a fixed budget, although reasoning overhead can sometimes outweigh computational savings; (ii) memory can reduce the cost of repeated solving but may preserve suboptimal strategies; (iii) compilation produces faster reusable solvers but at substantial upfront cost. Overall, computational algebra offers concrete, measurable targets for improving how agents discover and reuse structure.

cs.LG · cs.AI · cs.SC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.35890

Representational Simplicity and Circuit Size Dissociate in a Threshold-Dependent Way: A Controlled Test via Adversarial Training

Adam Elimadi

原文摘要

Sparse-autoencoder decomposability and concentrated feature attribution are increasingly treated as evidence that a model's computation is easier to reverse-engineer. Whether this representational and attributional cleanliness actually predicts a smaller or more tractable causal circuit remains an open question. We test this directly using adversarial training as a controlled instrument: it reliably reshapes internal representations, but this alone does not constitute a test of circuit size. We investigate this question through reverse-engineering complexity: the causal structure required to recover a model's behavior at a fixed level of faithfulness. To our knowledge, this is the first controlled empirical test of whether representational or attributional simplicity translates into causal simplicity at the circuit level. Starting from the same pretrained GPT-2 Small checkpoint, we apply matched standard and adversarial continual training, requiring both conditions to retain competence on indirect object identification and pass independent robustness verification before comparing mechanisms. We then compare the models along three complementary axes: sparse-autoencoder decomposability, SAE feature engagement in task attribution, and the size of faithful circuits recovered from the raw computational graph. The robust model is more SAE-decomposable and engages fewer SAE features in task attribution. Circuit size is regime-dependent: on competence-matched IOI, standard leads or ties below 85% faithfulness, but robust needs substantially fewer edges at high faithfulness (90%, 95%). The high-faithfulness advantage holds for every qualified robust model in a seven-point adversarial-strength sweep and for a second corpus, as do the representational trends.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16072

KVPath: Connecting Key-Value Representations Across Transformer Layers

Xuan Luo, Zhiyuan Wang, Hanzhong Cao, Xifeng Yan

原文摘要

Transformers typically cache an independent key-value (KV) representation for each token at every layer. Recent work reveals that a token's KV representations across neighboring layers are correlated and contain shared information. We hypothesize that this correlation implies incremental inter-layer changes, which can be effectively captured by low-dimensional deltas. Motivated by this, we propose KVPath, a new KV representation paradigm that connects KV representations across the network depth into a path. The path establishes a full-dimensional base KV at the first layer, while subsequent layers generate low-dimensional delta KVs to update the preceding KV, thereby constructing their own KV states. By caching the base KV once alongside these compact delta KVs, KVPath preserves distinct layer-specific representations without storing full KV states at every layer. We develop two variants of KVPath: KVPath-Residual, which updates vector values via residual addition, and KVPath-Rolling, which progressively updates feature dimensions in a rolling manner. Under matched parameter counts, KVPath reduces KV-cache storage by up to 68\% for Grouped-Query Attention (GQA) and 55\% for Multi-Head Latent Attention (MLA), while maintaining or even improving model performance.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16069

"It's Just a Small Weapon": What Rewarding Misbehavior During RL Does to Character-Trained Models

Mohammad Omar Khursheed, Thomas Morris, Fabien Roger

原文摘要

Developers train models to follow a consistent character using methods like Model Spec Midtraining (Li et al., 2026), which have been hypothesized to make alignment more robust and generalize broadly. However, it is possible for subsequent RL with misspecified rewards to work against character training. In this work, we apply character training to teach gpt-oss-20b a semi-realistic constitution and study how it interacts with two types of RL that reward breaking the constitution's rules (opposing RL): answering harmful requests rewritten as math problems, and reward hacking in agentic coding environments. The character-trained models learn both behaviors easily, though they are slightly slower than gpt-oss-20b to answer harmful requests. Character-trained models also display more motivated reasoning on harmful answers than gpt-oss-20b, and hack on some tasks which monitors find harder to catch. Finally, since explaining the reasons behind rules can improve how alignment generalizes (Kutasov et al., 2026), we compare constitutions that justify the same rules by grounding them in a value or not, and find little difference between motivations on misalignment, honesty, and capability evaluations.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16066

Optimal Additive Integer Masks for One-Guess Vulnerability under an Absolute-Moment Budget

Joonas Pääkkönen

原文摘要

We study the average one-guess success probability of an adversary who observes the ordinary integer sum of a finite-support secret and an independent integer-valued mask, subject to an absolute-moment budget on the mask. For a uniform secret on an interval the exact optimum and an explicit optimiser are obtained. For a biased bit the masks that achieve no increase over the prior baseline are characterised and their minimal cost is determined. The entire cost-vulnerability frontier is then determined for every bias. It is the piecewise-linear interpolation of an explicit sequence of points generated by cost-optimally shifted truncated geometric masks. It has infinitely many affine pieces, which accumulate at the critical cost, where the frontier meets the prior baseline with zero slope. The proof rests on a layer-cake decomposition of an arbitrary mask into truncated geometrics, along which both the vulnerability and the cost are exactly additive. The results give exact design guidance for additive integer noise under the average one-guess metric; they do not constitute a cryptographic masking scheme.

cs.IT · cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16065

Tokenized Reward Systems for Federated Learning: The Privacy vs. Fairness Tradeoff

Andreas Sendros, Nikolaos Pavlidis, Vasilios Perifanis, Pavlos S. Efraimidis

原文摘要

Federated learning (FL) systems increasingly rely on blockchain based reward mechanisms to incentivize participation while preserving data privacy. We design and empirically evaluate two complementary Solidity smart contracts for FL reward distribution: the first one that allocates rewards proportionally to a publicly disclosed quality score, and a privacy preserving variant that hides individual quality scores behind a zero-knowledge (ZK) proof of a threshold condition. Both contracts mint a non transferable, use-and-burn token maintaining a trustless incentive structure. We evaluate the two designs through MNIST training with adversarial behavior, including Sybil attacks, zero gradient and random noise, across three independent random seeds. Our results reveal a fundamental tradeoff between cryptographic privacy and incentive fairness. The privacy preserving architecture strictly degrades system robustness, doubling the adversary's average reward capture while introducing severe outcome variance. This vulnerability is driven by a structural self-inclusion bias within cosine-similarity scoring: as the global model converges, legitimate gradient updates naturally diminish, rendering them statistically indistinguishable from bounded noise. Adversaries exploit this dynamic to systematically breach the ZK threshold, triggering a late stage failure cascade that culminates in absolute adversary monopoly over the reward pool and the total exclusion of honest contributors. Alongside a closed-form theorem formalizing this privacy--Sybil resistance tradeoff, we map strict mainnet deployment boundaries via linear gas scaling models, explicitly incorporating the post-Fusaka EIP-7825 per transaction gas cap. Ultimately, while cryptographically secure, static threshold ZK gates fail as incentive layer fairness mechanisms under realistic FL convergence dynamics on standard benchmarks.

cs.LG · cs.AI · cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.16064

Privileged Information Does Not Guarantee Privileged Guidance: Auditing On-Policy Self-Distillation for Reasoning Models

Haoyi Qiu, Akshat Agarwal

原文摘要

On-policy self-distillation (OPSD) trains a student on next-token targets from a teacher, a frozen copy of the student that additionally receives privileged information, such as a worked solution, hidden from the student. Whether student gains reflect that information remains unclear. Taking the worked solution as privileged information, we audit four stages across three Qwen3 scales on mathematics: Source, whether the teacher can use the solution; Encoding, whether the resulting token targets are useful; Transmission, whether the student learns from the targets; and Expression, whether the student applies the learned behavior without the solution. With the solution, the teacher solves hard problems and recovers from student errors, yet the teacher does not reliably recognize errors, and its token targets do not favor correct trajectories. Repeated training makes successes within the student's reach more frequent rather than unlocking new ones, and corrections the student learns to predict are not reliably expressed in free generation. To isolate the solution's contribution, we vary its availability separately from teacher and student thinking modes. The largest gains occur when a teacher in thinking mode supervises a student trained and evaluated without an explicit thinking trace, and problem-only teachers often match or exceed privileged-solution teachers: the gain tracks the teacher's thinking mode rather than the solution, a pattern we call reasoning arbitrage. Accuracy gains alone therefore do not establish transfer of reasoning from privileged solutions: bottlenecks lie after Source, across Encoding, Transmission, and Expression. Future evaluations must measure what the solution adds to otherwise identical training.

cs.CL

查看论文
arXiv ↗PDF ↗