CS Paper Radar

Weekly top 20

2026-W46 Nov 09 – Nov 15, 2026

Ranked from 3,965 papers submitted on 7 days. Each research area gets slots in proportion to its share of this week's papers (at least one), filled by score; near-duplicates are removed. Why

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works
1

Discovering the Relation Between Architectural Models and Reference Architectures

cs.SE ·2026-11-10 ·Thi Dinh Tran, Maria Teresa Rossi, Ludovico Iovino, Leonardo Mariani
top venuecode/datahot direction
Abstract · Accepted at the 49th IEEE/ACM International Conference on Software Engineering (ICSE 2027). 13 pages, 5 figures, 4 tables. \c{opyright} 2027…

Assessing whether concrete software architectures are consistent with their reference architecture (RA) is essential to ensure that architectural decisions scale coherently across projects, preserving the quality attributes the RA was designed to guarantee. In the absence of dedicated techniques, mapping the components of an architectural model onto a predefined RA can be addressed only with general-purpose model-comparison algorithms. Our experiments show that these algorithms are inaccurate when applied to RAs. To address this gap, we propose Reference Architecture Mapping Discovery (RAMD), an LLM-based workflow that combines model analysis with structured prompting, and refines results through an iterative RA compliance-checking mechanism. We evaluate RAMD on 64 open-source architectural models spanning three RAs from the IoT, self-driving, and smart parking domains. With domain knowledge and GPT-4o, RAMD reaches 85.8% mapping accuracy and outperforms the strongest baseline by 44.6 percentage points, while ensuring structural consistency with the RA.

0.89
2

Hello Again: Exploiting Replay Protected Monotonic Counters to Brute-Force Windows Hello PINs

cs.CR ·2026-11-13 ·Christian Werling, Matti Finder, Hans Niklas Jacob, Julian Prein, Niclas Kühnapfel +1 more
top venue
Abstract · This is a slightly updated version of the paper published at ACM CCS '26 (https://doi.org/10.1145/3830454.3846631). The section Changes from…

Windows Hello is the primary authentication mechanism on over one billion Windows 11 devices, supporting PINs as short as four digits. Such low-entropy PINs are viable only because the TPM's anti-hammering protection mitigates brute-force attacks. Most modern x86 platforms rely on firmware TPMs (fTPMs) that persist their non-volatile (NV) state, including the lockout state, to SPI flash. On newer platforms, rollback prevention of this state is delegated to the flash chip via Replay Protected Monotonic Counters (RPMCs), a JEDEC-standardized mechanism operating over a physically accessible bus. We present the first practical security analysis of RPMC in shipping x86 systems. Using low-cost interposer hardware, we identify multiple implementation flaws, two of which fundamentally undermine RPMC's security model: First, Winbond SPI flash - the leading NOR flash vendor with over 25% market share - leaks its HMAC key under voltage glitching, enabling arbitrary counter forgery. Second, AMD's NV controller does not verify RPMC increments, so an interposer can suppress them unnoticed. Moreover, AMD platforms without RPMC flash appear to lack NV rollback protection altogether: on two AMD desktops, reflashing an old image resets the fTPM's lockout counters. Both flaws enable rollbacks of the fTPM's NV state (AMD, Intel, or Pluton), including its lockout counters. We demonstrate this end-to-end on locked, encrypted hardware: our setup finds a 4-digit Windows Hello PIN in 40 hours instead of 69 days. A compromised PIN also grants access to device-bound keys and Hello-based enterprise SSO. The same primitive defeats BitLocker's TPM+PIN pre-boot protector, reducing the worst-case time for a default-length 6-digit PIN from 19 years to 20 days.

0.83
3

Parsed, Plausible, Wrong: Output Format Controls Silent Errors in Small-Model Semantic Parsing

cs.CL cs.LG ·2026-11-12 ·Umar Ahmed, Maxim Dunaev, Mohammad Nour Al Awad, Sergey Ivanov, Ivan Khodnenko
benchmarkcode/data
Abstract · 10 pages, 6 figures, 11 tables. Code and models: https://github.com/UmarGit/coffee-llm

A small fine-tuned parser that emits a well-formed but wrong tree fails silently: nothing downstream can tell. We ask whether the output format, a free design choice, controls how often this happens. On the PIZZA ordering benchmark we fine-tune two sub-billion models (Qwen3.5-0.8B and LFM2.5-350M) with LoRA on six lossless encodings of the same target tree, from 44.3 down to 10.9 tokens per order, including two matched-length pairs that vary surface form at fixed length. Across 300 runs (5 training pool sizes, 5 seeds) with a hierarchical bootstrap and Holm correction fixed before the results were read, shortening alone never helps, but writing the target in ordinary words does. The word encoding is the most accurate format on both models (87.5% vs. 85.8% exact match for the native s-expression on Qwen) while emitting 4.1x fewer tokens, and it cuts the silent error rate by 2.3x on Qwen and 1.8x on LFM2, separating at every pool size. Error analysis locates much of the gain in negation. All runs also learn to refuse off-menu input (held out: 95.9-99.0% handled) and to apply in-utterance self-corrections. Strict refusal has a measured price: 2.1-10.1% of real, mostly polite, orders are refused.

0.82
4

Snoopie: A Multi-GPU Communication Profiler and Visualizer

cs.DC cs.PF ·2026-11-12 ·Mohammad Kefah Taha Issa, Muhammad Aditya Sasongko, Ilyas Turimbetov, Javid Baydamirli, Doğan Sağbili +1 more
top venuecode/datahighly cited
Abstract · Published at ICS '24" with the DOI 10.1145/3650200.3656597

With data movement becoming one of the most expensive bottlenecks in computing, the need for profiling tools to analyze communication becomes crucial for effectively scaling multi-GPU applications. While existing profiling tools including first-party software by GPU vendors are robust and excel at capturing compute operations within a single GPU, support for monitoring GPU-GPU data transfers and calls issued by communication libraries is currently inadequate. To fill these gaps, we introduce Snoopie, an instrumentation-based multi-GPU communication profiling tool built on NVBit, capable of tracking peer-to-peer transfers and GPU-centric communication library calls. To increase programmer productivity, Snoopie can attribute data movement to the source code line and the data objects involved. It comes with multiple visualization modes at varying granularities, from a coarse view of the data movement in the system as a whole to specific instructions and addresses. Our case studies demonstrate Snoopie's effectiveness in monitoring data movement, locating performance bugs in applications, and understanding concrete data transfers abstracted beneath communication libraries. The tool is publicly available at https://github.com/ParCoreLab/snoopie.

0.81
5

FocusBranch: Combinatorial Branch-and-Bound for $\ell_0$ Neural Network Robustness Verification

cs.LG ·2026-11-09 ·Renwei Deng, Yang He, Linyi Li, Yuepeng Wang
top venuecode/data
Abstract · Accepted to NeurIPS 2026. 32 pages, 11 figures, 6 tables

Verifying the local robustness of neural networks under $\ell_0$ perturbations is inherently combinatorial, as it must reason over all sparse subsets of $ε$ perturbed pixels. This paper presents FocusBranch, a combinatorial branch-and-bound framework for $\ell_0$ robustness verification. The key insight is to make the $\ell_0$-ball decomposable and branchable using a node abstraction, which consists of a focus pixel set $\mathcal{S}$ and a level $l$, permitting at most $ε-l$ pixels outside $\mathcal{S}$. This abstraction unifies two branching mechanisms: concrete nodes branch through upper shadows to fix additional perturbed locations, while super nodes branch through covering designs to cover many perturbation cases with fewer child nodes. FocusBranch proves node safety by concretizing linear relaxations over node-constrained $\ell_0$-balls and recursively branches on unknown nodes. Rather than relying on a fixed verification plan, it samples nodes across levels to estimate safe rates and costs, then selects a starting level and branching strategy predicted to minimize verification time. Experiments on fully connected and convolutional classifiers for MNIST, Fashion-MNIST, and CIFAR-10 show that FocusBranch verifies $\ell_0$ robustness effectively across architectures and outperforms state-of-the-art verifiers. The code is available at: https://github.com/simpl-group/FocusBranch.

0.80
6

From Link Prediction to Linear Community Detection: A Finite-Sample Guarantee for Graph Pretraining

cs.LG stat.ML ·2026-11-11 ·Ying Chen, Zhangyang Wang, Yixuan He
top venuetheorycode/data
Abstract · Accepted at NeurIPS 2026. 47 pages, 8 figures. Code: https://github.com/YChen1103/LP-to-Linear-Community-Detection

Link prediction (LP) is a standard pretraining objective for graph neural networks (GNNs): hide edges, train an encoder-decoder to predict them, and reuse the node embeddings for downstream node tasks. Despite its empirical success, there is little quantitative theory explaining why an edge-level objective should produce embeddings that are linearly useful for node classification. We study this question in the finite-sample stochastic block model (SBM). For any bounded GNN encoder paired with a symmetric bilinear decoder, we prove a conditional reduction: small population held-out LP excess risk implies small community-classification error for a transductive linear classifier on the learned embeddings. The bound decomposes into an optimization residual, an architectural approximation residual, and explicit statistical terms of order $\tilde{O}(\sqrt{d/n})$ in bounded dense regimes. We further give a constructive few-label classifier: nearest-centroid classification under the positive semidefinite quadratic form $G = W Z^\top Z W$, whose label error is coupled to LP quality. The framework is architecture-agnostic through the approximation residual, and we instantiate this residual as $O(\log n / n)$ for a graph isomorphism network (GIN) with spectral positional encoding and a bilinear decoder. Synthetic SBM experiments validate the theorem chain from LP risk to score error to classification error, and show when the worst-case constants become conservative.

0.77
7

Domain Adaptation Based Pipeline for Character Classification and Handwritten Text Recognition

cs.CV cs.LG ·2026-11-13 ·Florent Imbert, Simon Corbillé, Hui Han, Elisa H. Barney Smith
top venuecode/datahighly cited
Abstract

Character classification and handwritten text recognition, particularly in specialized tasks, often suffer from limited labeled data, while a large amount of unlabeled data remains unused. To address this challenge, we propose a novel unsupervised domain adaptation-based pipeline to reduce the distribution gap between source and target domains. Our pipeline follows three main steps. In the first step, we pretrained an expert model on source data, to learn a strong general representation. In the second step, we perform unsupervised domain adaptation to leverage unlabeled target data, adapting the model to the specific characteristics of the target domain. In the third step, we trained the expert model on source data in the target style. By aligning the feature distributions between the source and target, our method enables the model to generalize effectively across diverse datasets. The unsupervised training increases the computation time only during training not in inference. We evaluate the pipeline on both character-level and historical handwritten line recognition tasks, demonstrating its flexibility. We establish a new state-of-the-art on the Safran-MNIST-DLS dataset defined in Track 2 of the DAGECC competition with 76.17% Macro F1-score and achieve competitive results on each of the specific datasets of READ 2018 without using the labels of the target data. Code is available at https://github.com/simon-corbi/DRANet-for-classification-and-htr

0.76
8

SLICES: A Benchmark and Evaluation Framework for Self-Supervised Learning on Sparse ICU Time Series

cs.LG ·2026-11-13 ·Hannes Ill, Dimitrios Proios, Leo Anthony Celi, Peter J. Schüffler
top venuebenchmarkcode/data
Abstract · Accepted at the Machine Learning for Health (ML4H) Symposium 2026, Proceedings track (PMLR 346). 28 pages, 8 figures, 21 tables. v2: author …

Self-supervised learning (SSL) results for intensive care unit (ICU) time series are hard to compare when data processing, encoders, and adaptation vary together. We present SLICES, a benchmark and evaluation framework for comparing and extending SSL methods under a shared clinical setup. It fixes 24-hour value and observation-mask inputs, a Transformer encoder, patient-level splits, four endpoints, two adaptation protocols, and five matched seeds across MIMIC-IV, eICU-CRD, and their pooled union. Masked autoencoding (MAE), joint-embedding prediction (JEPA), and contrastive learning provide reference implementations. Among these objectives, MAE leads mean AUPRC in all nine frozen classification comparisons, also under converged probes, and in seven of nine fine-tuning comparisons. Fine-tuned MAE leads hospital mortality at full labels and scores higher than a supervised Transformer of the same architecture at every label fraction, and SSL fine-tuning gives the lowest length-of-stay error. Supervised baselines remain strong elsewhere: GRU-D leads 24-hour mortality and XGBoost leads acute kidney injury in all three regimes. Rankings depend on the adaptation protocol, the comparator set, and whether the probe loss matches the metric. A two-seed MIMIC-IV rerun with the maintained code stays within two across-seed SDs of the reference in all 48 downstream runs. The code repository supports new objectives, shared downstream evaluation, and regeneration of the reference results, so new SSL methods can be compared with these references under the same conditions. Generalization to unseen hospitals is not tested here and is the natural next extension.

0.74
9

Alloy: A Multi-Material Physics Engine for Robot Learning

cs.RO ·2026-11-12 ·Pranay Gupta, Deepanshu Rohilla
code/data
Abstract · 10 pages, 4 figures, 4 tables. Code: https://github.com/argonrobotics-ai/alloy

Robots handle much more than rigid objects: towels, cables, water, sand. The simulators used to train robots at scale are strongest on rigid bodies, and adding another material usually means a second engine and hand-written glue between the two. We present Alloy, an open-source physics engine for robot learning in which rigid robots, cloth, cables, liquids, granular material and soft bodies share one scene. Each object is described once; Alloy assigns it to a solver suited to its material, couples the solvers inside every control step, and runs the scene in many parallel worlds on one GPU, or on a CPU. Every world carries a health grade, to help tell a failure of the physics from a failure of the policy. Alloy is built on Newton, MuJoCo Warp and Warp, and it is written to be driven by coding agents: its documentation ships inside the package, every error names its fix, and the engine is exposed as tools over the Model Context Protocol. In a cold test, agents that had never seen Alloy installed a pre-release from the package index and trained policies for walking, cable handling and pouring on a single GPU. We also report how throughput grows with the number of worlds, depth-camera throughput, agreement with closed-form references, and a reliability test over a thousand worlds.

0.73
10

When Expert Disagreement Hurts: Prestige and Selectivity in LLM Decision Revision

cs.AI cs.CL ·2026-11-12 ·Yupeng Tang, Mingfeng Lin
top venuecode/data
Abstract · Accepted at NeurIPS 2026. 38 pages, 7 figures, 36 tables. Code and data: https://github.com/yupengtang/when-expert-disagreement-hurts

When language models revise a decision after another agent disagrees, neither reversal counts nor final accuracy shows whether they are fixing mistakes or abandoning correct answers. We introduce a controlled two-pass audit that holds the question, the evidence, and the model's initial answer fixed and varies only the revision context: neutral reflection, anonymous disagreement, or disagreement attributed to a domain expert. For binary decisions the accounting is exact: revision raises accuracy only if it is selective enough, reversing wrong answers more than $a/(1-a)$ times as often as correct ones, where $a$ is initial accuracy (nine times at 90%). Across five models and 1,989 evidence-grounded decisions in medicine, science, and law, no model clears this bar under expert disagreement, and accuracy falls by up to 16.1 points. The expert label, a prestige cue, adds reversals for four of the five models, yet its effect on accuracy ranges from a 5.2-point loss to a 2.4-point gain, so reversal counts are a poor proxy for harm. With identical arguments from both sources, the label still raises reversal for three proprietary models and in a pooled analysis of four newer models. Models also respond to stated source reliability. Revision pipelines should therefore be evaluated with paired, conditional metrics, per model and protocol.

0.73
11

TorchCurves: Differentiable Parametric Curves

cs.LG stat.ML ·2026-11-10 ·Alex Shtoff
code/data
Abstract · Software paper for a library I've been working on for more than a year

We present TorchCurves, a PyTorch library that treats vector-valued parametric curves as learnable primitives. Its central operation maps batches of scalar inputs to points on curves represented in B-spline or Legendre polynomial bases, with differentiation through both inputs and coefficients. Curve coefficients can be optimized jointly with the surrounding model or produced by another network. The library provides trainable modules, reusable basis modules, and a functional interface, supporting applications ranging from continuous feature embeddings to Kolmogorov-Arnold networks and functional regression. Its implementation combines classical evaluation algorithms with vectorized PyTorch operations, emphasizing readability and maintainability without custom hardware kernels. Timing experiments compare curve evaluation against NumPy and SciPy and KAN training against EfficientKAN across a range of problem sizes. The results illustrate where this generality remains computationally practical and where overhead is substantial. Library is available on GitHub at https://github.com/alexshtf/torchcurves.

0.72
12

Factored One-sided Adam-Moment for Practical and Scalable SOAP

cs.LG ·2026-11-12 ·Jaemyung Yu, Byeongho Heo, Sangdoo Yun, Dongyoon Han
publishedcode/data
Abstract · NeurIPS 2026. 17 pages, 5 figures, 11 tables. Code: https://github.com/naver-ai/foam

Second-order optimizers (e.g., Shampoo, SOAP) have emerged as compelling alternatives to Adam/AdamW and promise faster per-step convergence than AdamW in transformer training. However, they still remain costly in practice, and even SOAP still pays substantial optimizer-state memory and eigendecomposition overhead. We ask whether the benefits of SOAP can be retained at nearly the efficiency of AdamW. In this paper, we answer yes with a lightweight SOAP variant, Factored One-sided Adam-Moment (FOAM). The path to practical impact is twofold: first, FOAM employs input-only preconditioning, dropping the output-side Kronecker factor to reduce memory and remove one eigendecomposition. The Kronecker structure of the Hessian shows that this is exact for a diagonal output-side curvature and nearly lossless for a near-diagonal one, as we observe empirically. Second, FOAM replaces SOAP's dense rotated second moment with a factored Adam-moment estimator based on row-by-column statistics. Under the K-FAC Fisher assumption with a diagonal output curvature, this estimator exactly recovers SOAP's dense second moment, and we measure only a 15-20% K-FAC violation, with negligible effect on the update. Our empirical results support the claim that FOAM matches strong SOAP and Muon baselines in validation loss across scales of large language model (LLM) pre-training, uses half of SOAP's optimizer state, runs at Muon-class wall-clock time, and retains robustness across a three-fold learning-rate range. The results hold from 124M to 760M parameters, demonstrating that SOAP's gains need not require SOAP's cost. Code: https://github.com/naver-ai/foam

0.71
13

AutoVLA-D: Learn a Trajectory-Geometric Condition from Reasoning for Autonomous Driving VLA

cs.RO cs.CV ·2026-11-09 ·Yuguang Yang, Zhewen Tan, Canyu Chen, Linlin Yang, Cheng Chi +7 more
top venuecode/data
Abstract · Accepted by NeurIPS 2026

Autoregressive VLA planning usually predicts continuous trajectories as the final output, but driving instructions often describe actions rather than ego states, which weakens direct conditioning on the final continuous trajectory. We address this gap with TrajCond, which introduces a discrete trajectory-geometric condition for VLA planning. Specifically, we represent TrajCond as a 1D-Brohan differential action trajectory and learn it with a two-stage training framework before final continuous trajectory prediction. With the simultaneous autoregressive generation of discrete and numeric trajectories, we name our method AutoVLA-D (Double). Experiments on NAVSIM navtest, nuScenes, and NeuroNCAP show that discrete trajectories are weaker as direct outputs but more effective as TrajCond, consistently improving final continuous trajectory prediction across closed-loop and open-loop driving benchmarks. Code is available at https://github.com/ShivaTeam/AutoVLA-D.

0.70
14

Heavy Metal Acrobatics: Learning Dynamic Control for Skid-Steer Loaders

cs.RO ·2026-11-13 ·Pierre-Jean Lorentz, Pascal Egli, Edo Jelavic, Cliff Li, Ramitha Sundar +7 more
top venue
Abstract

Automation of heavy machinery offers the opportunity to improve productivity in mechanized industries such as construction, mining, and material handling. Yet, automation of such machines is challenged by their complex dynamics, in part because of hydraulic actuation that involves fluids, flexible elements, and nonlinear delays. A concomitant challenge is the size and weight of these machines, which can make experiments on the physical systems exceedingly dangerous. Here, we show that \ac{RL} in simulation, followed by transfer of the learned policies from simulation to reality, can successfully automate acrobatic maneuvers that push a heavy construction machine to the limits of its actuation and have only previously been accomplished by a small number of elite human operators in official world-record performances. We trained policies that perform autonomous acrobatic maneuvers on a John Deere 332G, an operational 5-tonne skid-steer loader. The policies can autonomously swing the machine up onto two wheels and, while balancing on two wheels, drive over rough terrain, trace out specified trajectories, spin around, and swing the bucket back and forth before landing it safely back onto four wheels. This work demonstrates that \ac{RL} and sim-to-real transfer enable dynamic autonomous control of heavy machinery, directly contributing to greater automation in heavy equipment sectors. Videos: https://sites.google.com/leggedrobotics.com/heavy-metal-acrobatics This is the author's version of the work. It is posted here by permission of the AAAS for personal use, not for redistribution. The definitive version was published in Science Robotics on 11 November 2026, DOI: https://doi.org/10.1126/scirobotics.aef0894

0.70
15

"God says we are right!": The Interplay between Religion and Propaganda on Arabic Social Media

cs.SI ·2026-11-15 ·Mahmoud Fawzi, Björn Ross, Walid Magdy
top venue
Abstract · published at CHI 2026 @inproceedings{fawzi2026god, title={''God says we are right!'': The Interplay between Religion and Propaganda on Arabi…

Religion is a main aspect of life in many parts of the world; hence, it acts as a powerful tool for influencing people's views and actions. In the context of propaganda and misinformation, religion has been perceived as a factor that impacts people negatively leading them to be influenced by false or biased information. However, there are limited quantitative studies that explore the exact role of religion in this process. In this study, we investigate whether state-sponsored propaganda accounts share religious content in a different way than the typical norms reported in the literature and whether they mobilize this content to promote their agendas. This is achieved by exploring 15 Middle Eastern propaganda Twitter datasets encompassing around 124 million tweets from 32.5K accounts attributed to KSA, UAE, Egypt, Turkey, and Iran. We find that propagandists share excerpts from the Quran and Hadith according to patterns that reflect aspects of their state's agenda. We also demonstrate examples where such religious texts were used to modulate political messages.

0.68
16

Pair-Merge Tokenisation Is Hard to Approximate on Binary Strings and Strongly NP-Hard on Unary Datasets

cs.DS cs.CCcs.CL ·2026-11-10 ·Kundan Baliga, Aiden Xie, Aditya Byju
theory
Abstract · 34 pages, 11 figures. Code and certificates: https://doi.org/10.5281/zenodo.24023730

Tokenisers such as Byte-Pair Encoding (BPE) shorten text by repeatedly merging an adjacent pair of symbols into a new one; fewer tokens mean less computation. We ask whether an efficient algorithm can choose $k$ merges that leave a string of length $n$ almost as short as possible; doing nothing is already within a factor $n$. We prove that, unless P = NP, no polynomial-time algorithm comes within a factor $c\,n/\log n$, for some constant $c>0$, already for one binary string and for BPE-style merges that replace every occurrence. The key idea is that $k$ merges form a grammar with $k$ rules: a vertex-cover gap becomes a grammar-size gap, which we amplify with relabelled copies, encode in binary and realise with explicit merges. For merges that may skip occurrences, a simple algorithm achieves $O(n/\log\log n)$, and no Karp reduction can prove hardness beyond order $n/\log n$ unless coNP $\subseteq$ NP/poly. Over a one-letter alphabet, both tokenisation problems left open by Kastreva et al. are strongly NP-complete on datasets with repeated strings.

0.65
17

Training Vision Transformers to Focus: Emergence of Spatial Attention Head Specialization

cs.CV ·2026-11-09 ·Yuji Nozawa, Yu-Chieh Lin, Youyang Ng
top venuecode/data
Abstract · Accepted at NeurIPS 2026. Project page: https://yuji-nozawa.github.io/ReFoCo/

Self-supervised learning in Vision Transformer models, such as DINO, enables robust visual perception without requiring labeled data by leveraging large-scale image datasets. However, these models lack a focusing mechanism during training, resulting in perceptual representations that are often ambiguous or overly similar. In this work, we introduce two visual focusing loss functions designed to establish correspondence between the model's attention and specific regions within images. Specifically, we leverage the multi-head attention mechanism within the model to selectively steer the model's focus, leading to the emergence of spatial attention head specialization, characterized by diverse and spatially focused perceptual capabilities without requiring annotations. Through both qualitative and quantitative evaluations, we demonstrate that our method substantially increases the spatial selectivity and diversity of attention heads, improving both interpretability and performance on dense prediction tasks such as zero-shot image segmentation.

0.61
18

Specify Privacy Yourself: Assessing Inference-Time Personalized Privacy Preservation Ability of Large Vision-Language Models

cs.CV cs.AIcs.CLcs.CY+1 ·2026-11-10 ·Xingqi Wang, Xiaoyuan Yi, Xing Xie, Jia Jia
top venuepositioncode/data
Abstract · Accepted by ACM Multimedia 2025 (Brave New Ideas track)

Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities but raise significant privacy concerns due to their abilities to infer sensitive personal information from images with high precision. While current LVLMs are relatively well aligned to protect universal privacy, e.g., credit card data, we argue that privacy is inherently personalized and context-dependent. This work pivots towards a novel task: can LVLMs achieve Inference-Time Personalized Privacy Protection (ITP$^3$), allowing users to dynamically specify privacy boundaries through language specifications? To this end, we present SPY-Bench, the first systematic assessment of ITP$^3$ ability, which comprises (1) 32,700 unique samples with image-question pairs and personalized privacy instructions across 67 categories and 24 real-world scenarios, and (2) novel metrics grounded in user specifications and context awareness. Benchmarking the ITP$^3$ ability of 21 SOTA LVLMs, we reveal that: (i) most models, even the top-performing o4-mini, perform poorly, with only $\sim$24% compliance accuracy; (ii) they show quite limited contextual privacy understanding capability. Therefore, we implemented initial ITP$^3$ alignment methods, including a novel Noise Contrastive Alignment variant which achieves 96.88% accuracy while maintaining reasonable general performance. These results mark an initial step towards the ethical deployment of more controllable LVLMs. Code and data are available at https://github.com/achernarwang/specify-privacy-yourself

0.61
19

khammoon: From a Linguistic Guideline to a Thai Morpheme Segmenter

cs.CL ·2026-11-09 ·Warit Sirichotedumrong, Adisai Na-Thalang, Chanakan Wittayasakpan, Kritsadha Phatcharoen, Kunat Pipatanakul
code/data
Abstract · Technical Report. Code, models, and data are available at https://github.com/warit-s/khammoon-segmenter

Thai is written without spaces between words, making segmentation a foundational step in most Thai NLP pipelines. Existing tools primarily reproduce the word-level conventions of corpora such as BEST; however, because the notion of a "word" in Thai lacks a unified linguistic consensus, these conventions struggle with productive compounding, leading to arbitrary boundaries and ambiguity in downstream representations. Instead, we propose operating at the morphological level, focusing on khammoon, the smallest atomic unit that carries meaning. Segmenting at the morphemic level provides an explicit, consistent morphological representation that eliminates compounding ambiguity. Crucially, this representation benefits downstream learning: models trained on morphemes perform on par with or exceed word-level baselines across part-of-speech tagging, named-entity recognition, and sentiment classification, demonstrating that linguistic consistency improves representation quality without downstream performance penalty. To operationalize this representation, we translate linguistic principles into a formal decision guideline with near-perfect human agreement ($κ\approx 0.99$), scale annotation with an LLM, and correct systematic errors with human labels. Finally, we train khammoon, a lightweight 1M-parameter convolutional BiLSTM segmenter designed for efficient CPU inference, achieving a boundary F1 of 0.992 on out-of-domain text. The guideline, datasets, model, and library are all publicly released.

0.60
20

Is Spoken Hungarian Low-resource?: A Quantitative Survey of Hungarian Speech Data Sets

cs.CL ·2026-11-11 ·Péter Mihajlik, Katalin Mády, Anna Kohári, Fruzsina Sára Vargha, Gábor Kiss +2 more
publishedsurveycode/datahighly cited
Abstract · LREC 2024

Even though various speech data sets are available in Hungarian, there is a lack of a general overview about their types and sizes. To fill in this gap, we provide a survey of available data sets in spoken Hungarian in five categories (e.g., monolingual, Hungarian part of multilingual, pathological, child-related and dialectal collections). In total, the estimated size of available data is about 2800 hours (across 7500 speakers) and it represents a rich spoken language diversity. However, the distribution of the data and its alignment to real-life (e.g. speech recognition) tasks is far from optimal indicating the need for additional larger-scale natural language speech data sets. Our survey presents an overview of available data sets for Hungarian explaining their strengths and weaknesses which is useful for researchers working on Hungarian across disciplines. In addition, our survey serves as a starting point towards a unified foundational speech model specific to Hungarian.

0.60