Universes are central to dependent type theory, and they are notoriously difficult to handle in a way that is both correct and usable. We propose a new "fuss-free" generalised algebraic presentation for polymorphic cumulative universes that dispenses with the intricate theory of coherent universe coercions in favour of a simpler formulation, which we prove equivalent by means of a normalisation theorem for the former. We study the fuss-free formulation by means of (1) an abstract specification of its bidirectional elaboration algorithm, and (2) a concrete implementation in Haskell. We also describe and implement an extension of the fuss-free universe hierarchy with a judgemental notion of datatype description from which prior notions of cumulative inductive type may be derived.
Learned sparse embeddings are a powerful alternative to dense retrieval, yet they have fallen behind single-vector and late interaction models. One reason for their limited adoption is that they are notoriously difficult to train, generally requiring very tight conditions to converge. We present a recipe for training Splade-like encoders on modern backbones and data that achieves a better efficiency-effectiveness trade-off. We discuss the role of logit distribution and its impact on the training stability and the final sparsity of the encoder. We also introduce a new per-position fan-out control - a hard cap on how many vocabulary dimensions each position may emit. We further apply a simple vocabulary merging strategy that removes redundant entries - such as case and word-boundary variants of the same word - which are common in cased BPE-based backbones. We apply our recipe to the LateOn backbone and training data, obtaining SPARSEUP, a sparse encoder that reaches 56.4 average nDCG@10 on BEIR-13, the highest we know of among public vocabulary-based sparse encoders below 150M parameters.
Chang Rajani, Luke Sebastien Marsden, Stephan Kunz, Claudio Mucignat, Ivan Lunati
原文摘要
Particle image velocimetry (PIV) reconstructs a flow field from the motion of tracer particles but the reconstruction quality is bounded by how well the flow structure of interest is resolved in the images. The acquisition position is typically fixed by the operator before the measurement. We cast camera placement as a sequential decision-making problem and train a reinforcement learning agent to control a 6-DoF manipulator equipped with a camera, enabling it to locate, frame and resolve a hydrofoil wake in a water tunnel. The agent commands the camera pose and observes the scene only through the velocity field and uncertainty estimated from time-separated image pairs. It is trained entirely in a photorealistic simulation of our water tunnel, with wakes generated by Computational Fluid Dynamics (CFD) simulations of hydrofoils with varying profiles, positions and angles of attack. After zero-shot transfer to the real water tunnel, the policy reliably locates and frames the wakes of randomly placed hydrofoils. It resolves the wake at a higher resolution than a fixed camera view and with lower error in the measured flow statistics, in roughly half the time of a full scan and with far fewer image pairs.
Traffic counts are routinely used to calibrate travel demand models, particularly when the origin-destination (OD) matrix is not observed directly. An important modelling decision is whether to adjust individual OD flows or the parameters of the demand model that generates them. The latter approach uses fewer parameters and preserves a relationship between zonal characteristics and demand, but its accuracy depends on the assumed model structure. We compare three established approaches using the same prior information, traffic counts and equilibrium assignment. They estimate five production-gravity coefficients, origin and destination adjustment factors, or individual OD flows. Synthetic experiments on the Eastern Massachusetts and Chicago Sketch networks allow recovery to be assessed against known demand under different forms of model misspecification. Calibrating the five-coefficient model gives better OD recovery and held-out link-flow prediction when demand follows its assumptions or a two-purpose gravity process. Its advantage declines or reverses when destination choice or production relationships vary across zones. Cellwise adjustment achieves the closest fit to calibration counts in 65 of 66 evaluated conditions, yet it recovers OD demand least accurately of the three approaches in 45 of them. These findings show that model parameterisation matters beyond calibration fit. For transport planning, traffic counts should be used alongside independent evidence on trip production and distribution when selecting and validating a demand model.
Authentication solutions are the key design, architectural, and implementation decisions that implement authentication security controls. An example is a password-based authentication solution. While researchers often examine authentication quality, existing approaches do not distinguish which effects on quality stem from which aspect of authentication. For example, multi-factor authentication solutions tend to be less usable because they combine multiple authenticators, whereas passkeys are more secure than passwords because of properties inherent to the authenticator itself. This paper presents AQUA, a modular quality modeling framework for authentication solutions. AQUA's core idea is to decompose authentication into four interrelated concepts, enabling modular specification of authentication quality and tracing quality effects back to their source. We demonstrate AQUA with initial definitional quality models that address the levels of authenticators, authentication techniques, and authentication solutions. We further illustrate the interrelations between these concepts through concrete examples and a usability metric, showing how quality propagates across the concepts and ultimately affects the overall quality of an authentication solution. Our results provide a foundation for future work, including designing assessment and prediction models for authentication solutions and automated quality assessments.
Qishen Wang, Pengxiang Ding, Shifeng Bao, Mingyang Sun, Hengtao Li, Yi Gao, Baohua Yin, Yang Liu 等 17 位作者
原文摘要
Vision-language-action (VLA) models typically generate actions in chunks, organizing continuous robot execution into separately predicted segments and thus introducing potential discontinuities at action-chunk transitions. We propose ODEAction, a chunk-free action generation paradigm that replaces action-chunk prediction with Continuous Action Evolution. Given visual observations, language instructions, and robot proprioceptive information, ODEAction learns an action transition field defined over action space, where a transition vector characterizes the evolution rate of the current action with respect to evolution time. During execution, the model takes the current action as its initial condition and integrates the learned field using an ordinary differential equation (ODE), recursively generating subsequent actions with explicit dependencies. ODEAction is implemented as an action head and does not depend on a specific vision-language backbone architecture. Experiments on LIBERO, EBench, RoboTwin, and real robots show that ODEAction improves both task performance and trajectory quality over matched action-chunking baselines, with particularly pronounced gains on fine manipulation tasks.
Joshua Stiller, Santo M. A. R. Thies, Felix Czaja, Eyke Hüllermeier
原文摘要
Understanding model predictions is essential for physical applications, where outputs often inform safety-critical decisions, such as structural load assessment, weather warnings, and clinical diagnosis. Shapley values satisfy many desirable properties as an attribution method, but their computational cost during inference hinders their practical use. Current amortized explainers, such as FastSHAP, are limited to homogeneous inputs, which is problematic for physical applications where data often comes from irregular grids and geometries. We introduce OperatorSHAP, a grid-agnostic attribution method that allows us to train FastSHAP-like explainers for neural operators. We establish a theoretical framework for attributions in function space, connecting to Aumann-Shapley values. We further show that OperatorSHAP's explanations are consistent with state-of-the-art discrete Shapley values across resolutions and transfer across grid sizes without retraining.
Kevin Glocker, Kätriin Kukk, Romina Oji, Marcel Bollmann, Marco Kuhlmann, Jenny Kunz
原文摘要
Training high-performing language models which include medium- and lower-resource languages remains a challenge. Massively multilingual models still underperform language-specific adaptations, especially at smaller scales, due to the curse of multilinguality. In this work, we investigate scaling as an efficient strategy for adapting pretrained models to new target languages. Through comprehensive ablations matched for data and training FLOPs, we test whether upscaling from English base models enables more effective and resource-efficient adaptation than standard continued pretraining. We find that upscaling is an efficient strategy for improving target-language performance by relieving capacity constraints, especially for higher-resourced languages. An initial upscaling step in English is beneficial for low-resource languages. However, in higher-resource scenarios it is more compute-efficient to directly scale up on the target language, albeit slightly compromising data efficiency for most languages. For multilingual continued pre-training, we identify directly scaling up using the target data as the optimal setup for both data efficiency and reduced English forgetting.
Robots increasingly choose actions by searching against a reward model learned from human preferences, in model-predictive control or best-of-$N$ selection; we measure what this does to the true objective. On HalfCheetah and Hopper with crowd-sourced labels, and on Meta-World drawer-open and plate-slide with synthetic labels and blind labels from the two authors, harder MPPI search cuts true return by up to 76%: with human labels the drawer opens in 0 of 20 runs against 10 of 10 for the cloned policy the planner starts from, while held-out label accuracy stays at 0.75-0.97. On plate-slide human labels are exploited less than synthetic ones, and on window-open search helps. True return roughly follows how far the planner's weights move from its proposals ($\sqrt{\mathrm{KL}}$) for both weighting rules we test, and best-of-$N$ with a verifier from synthetic labels falls below a single noisy action. We prove that $χ^2$-regularized MPPI at a temperature set by the scorer's error spread on the proposals is worst-case optimal, that no positive gain can be certified once that spread exceeds the scorer's own, and that an episode's return splits exactly into per-step terms, of which the guarantee covers the selection terms. Calibrated from 100 proposal pairs ranked by true return (an oracle) on start states the evaluation never uses, this rule is the safest fix we test: with each of five independent sets it succeeds in every arm run and keeps on average 86-94% of the cloned policy's return on plate-slide and drawer-open, though it is cautious where search helps. Calibration from a person's comparisons cannot detect a bias the scorer learned from that person; correcting for the pairs people cannot tell apart, or a second person, limits the damage.
Raphaël Bonnet-Guerrini, Johann Ioannou-Nikolaides, Inar Timiryasov, Vincenzo Piuri
原文摘要
We present a first application of sparse-autoencoder-based mechanistic interpretability to particle physics. Studying a neutrino foundation model pretrained on IceCube data and fine-tuned for direction reconstruction, we identify a validated atlas of physical concepts in the model representation, using a strict validation protocol consisting of held-out tests, matched nuisance controls, and replication across independent dictionary trainings. Causal interventions show that the direction head barely draws on this atlas. Motivated by this underused information, we train an uncertainty head on the same event-level representation to predict the model's angular reconstruction error. Unlike the direction head, it depends causally on quality and brightness features from the atlas. These results suggest that mechanistic interpretability can reveal learned latent physics encoded within a model's internal representation and help design downstream tasks that exploit it.
Measurement exploration reveals how interacting local controllers behave but can alter the steady-state regime a network operator seeks to optimize. Exploratory rewards may therefore fail to identify a configuration's no-exploration value: under full exploration, identical logged distributions can conceal different target values. Identification requires accounting for both action-law mismatch and shifts in the steady-state context distribution. We introduce CAUSAL-UCB, which selects configurations using optimistic bounds on target causal values. Its interference-aware clipped doubly robust evaluator produces predecision certificates separating statistical variation from calibrated identification, transport, and temporal-dependence errors. For any finite horizon and predetermined schedule of measurement-exploration rates $\{η_t\}$, we prove lower bounds of $Ω((1+\sum_tη_t)^{-1/2})$ on minimax inference error and $Ω(\sum_tη_t)$ on measurement-induced deployment loss in a stationary, potentially slow-mixing family. We bound regret under valid certificates and selected-information growth, and establish an anytime finite-sample certificate with a shared radius for estimation, selection, and reporting in a restricted known-local-interference model. In a high-mismatch benchmark with 24 paired seeds and 200 rounds, CAUSAL-UCB attains target regret of $1.732$ versus $6.326$ for certified IPS-UCB, with deployment gaps of $30.607$ and $23.217$, respectively. Independent-audit coverage tests and trace-driven semi-synthetic studies further assess certificate coverage and deployment trade-offs.
In this paper, a 3D backbone network (SpconvTransformer) based on the fusion of convolutional architecture branch and Transformer architecture branch is proposed. The convolution branch uses 3D sparse convolution and submanifold convolution to learn local voxel features and semantic information, and the Transformer branch uses the self-attention mechanism to model the remote relationship between voxels, uses the broad receptive field to capture rich context information, and complementing the advantages of the two architectures by fusing the local features extracted from the two branches and the global features solves the problem of limited receptive fields of 3D convolutional neural networks. The experimental results show that the average accuracy of SpconvTransformer in 3D object detection and bird's-eye view detection tasks reaches 81.45% and 88.09%, respectively, and the average directional similarity reaches 91.10%, which is significantly better than other existing algorithms.
Yeoh Wei Zhu, Naresh Goud Boddu, Yao Ma, Shaltiel Eloul, Giulio Golinelli, Yash Satsangi, Rob Otter, Kaushik Chakraborty
原文摘要
Traditional financial institutions face inefficiencies that can be addressed by distributed ledger technology. However, a primary barrier to adoption is the privacy concerns surrounding publicly available transaction data. Existing private protocols for distributed ledger that focus on the Ring-CT model are not suitable for adoption for financial institutions. We propose a post-quantum, lattice-based transaction scheme for encrypted ledgers which better aligns with institutions' requirements for confidentiality and audit-ability. The construction leverages various zero-knowledge proof techniques, and introduces a new method for equating two commitment messages, without the capability to open one of the commitment during the re-commitment. Subsequently, we build a publicly verifiable transaction scheme that is efficient for single or multi-assets, by introducing a new compact range-proof. We then provide a security analysis of it. The techniques used and the proofs constructed could be of independent interest.
When researchers refine ideas with large language models over several turns, do the models preserve the original brief's constraints? We introduce DriftBench, a benchmark for constraint adherence in multi-turn LLM-assisted scientific ideation. It pairs 38 research briefs from 24 scientific domains, each with hard constraints and banned moves, with four interaction conditions and seven models from five providers (two open-weight), scored on 2,116 runs. Iterative pressure to improve a proposal raises structural complexity in every model and lowers adherence to the brief's hard constraints. A restatement probe shows that the constraints stay legible and reproducible on demand while the proposal departs from them. The knows-but-violates (KBV) rate, the share of pressure runs scored non-compliant on a hard constraint despite a correct restatement, exceeds 50% for most models. A crossed-judge design measures the judge offset. GPT-5.4 scores adherence 0.535 points lower than Opus 4.6 on the same runs. The design finds no judge-by-condition interaction on adherence, so the pressure effect on adherence is judge-stable while KBV levels and the bottom of the model ranking are not. Re-eliciting the brief at checkpoints lowers KBV for some models but not others, and complexity inflation persists. Blind human raters find more constraint violations than the LLM judge, so reported adherence is conservative. We release all briefs, prompts, rubrics, transcripts, and scores.
At the heart of a language model agent is an orchestrator program which is responsible for the state transition between consecutive turns. In existing systems, this program is implemented in the harness. This paper introduces self-programmed execution (SPE), in which the model completion is itself the orchestrator program, and the harness evaluates this program but does not impose any fixed orchestration policy. Formally, an SPE state of an agentic machine is one from which a model completion can load any successor state up to an embedding, and an instance of an SPE state is one which executes the model completion as a program. In practice, the program requires a language with unusual features, as its source code doubles as the content of the model's context window. I therefore introduce Spell, a Lisp dialect in which programs can edit and re-evaluate themselves, and effectful expressions like model invocations are structured such that re-evaluating an edited program does not replay its side effects. Experiments with existing models, not trained for SPE or Spell, show that frontier models can learn Spell in-context and accomplish challenging agentic tasks. These results demonstrate how an LM can act as an agent without any fixed orchestration policy, and they raise the question of what self-orchestration strategies might be learned by a model trained for self-programmed execution. Code is available at https://github.com/lukejoconnor/spell.
Modern distributed networks, notably transformers, acquire a remarkable ability (termed `in-context learning') to adapt their computation to input statistics, such that a fixed network can be applied to data from a broad range of systems. Here, we provide a complete mechanistic characterization of this behavior in transformers trained on a finite set $S$ of discrete Markov chains. The transformer displays four algorithmic phases, characterized by whether the network memorizes and generalizes, and whether it uses 1-point or 2-point statistics. We show that the four phases are implemented by multi-layer subcircuits that exemplify two qualitatively distinct mechanisms for implementing context-adaptive computations. Minimal models isolate the key features of both motifs. Memorization and generalization phases are delineated by two boundaries that depend on data diversity, $K = |S|$. The first ($K_1^\ast$) is set by a kinetic competition between subcircuits and the second ($K_2^\ast$) is set by a representational bottleneck. A symmetry-constrained theory of a transformer's training dynamics explains the sharp transition from 1-point to 2-point generalization and identifies key features of the loss landscape that allow the network to generalize. Put together, we show that transformers develop distinct subcircuits to implement in-context learning and identify conditions that favor certain mechanisms over others.
How does using less conversational history change reader-token use and benchmark scores? We examine this question on 1,100 BEAM questions. We held the retrieval ranking and reader constant across five fixed passage budgets, and mean assigned scores rose as the passage budget grew. The score gain per additional token became smaller at the upper tested budgets. Answerable questions generally received higher scores with more context. Pooled abstention scores sometimes rose as the passage budget shrank. We also tested message retention in a controlled task with explicit conversational-state targets and compared retained user messages with token-matched recency. Qwen matched the full state more often for direct requirements when user messages were retained. Its full-state matching declined when answers depended on references to earlier offers. Mistral produced near-floor results across the retention conditions. A scoring audit leaves the correctness of BEAM answers unresolved. These experiments characterize the token and score consequences of the tested memory budgets. Reliable budget selection depends on retaining the evidence each question requires and validating the measure used to assess answer quality.
Online 3D street-scene asset construction turns multi-view video into a continually updated scene of independently editable objects. The central challenge is to accumulate partial, intermittent observations under motion and occlusion without fragmenting identities or mixing evidence across objects. To address this challenge, we propose Video2Asset, an online framework for constructing persistent 3D assets from multi-view video. Using occupancy as an intermediate representation, we construct identity-aware observations with Historical Carrier Guidance for instance decoding and an identity encoder fusing instance-aligned 3D features with multi-view appearance. Accumulated-Evidence Association compares current observations with identity, geometric, and appearance evidence accumulated within asset records to determine identity inheritance and update world-coordinate states. We instantiate meshes from an asset library using these states and combine them with an occupancy-derived static background to support scene visualization, mesh replacement, object deletion, and trajectory editing. Experiments on several datasets demonstrate improvements in temporal identity consistency and asset quality across multiple occupancy models, including a 15.4% reduction in asset fragmentation on nuScenes with SGE.
Mateusz Meller, Joseph C. R. Thacker, Thomas W. Keal, Vendel Szeremi
原文摘要
In this work, we focus on addressing challenges related to modelling the integration of conventional high-performance computing (HPC) and quantum computing (QC). We introduce the Quantum Assembly Compiler (QASC) for OpenQASM3.0 files. The compiler produces an executable that leverages the high-performance simulator QuEST to simulate quantum programs. Additionally, we introduce a performance modelling tool, QACC-perf, which can estimate the cost of integrating QC in classical HPC workflows, as well as the execution time of the quantum circuits. Finally, we demonstrate the utility of these tools in the context of multiscale quantum chemistry workflows using the ChemShell computational chemistry environment.
Graph Retrieval-Augmented Generation (GraphRAG) connects information across documents through entity relationships to support relational reasoning. However, black-box access to GraphRAG systems can enable adversaries to reconstruct knowledge graphs and recover both structural connectivity between entities and the semantics of their relationships. Existing defenses lack formal privacy guarantees and provide limited protection against such reconstruction attacks. We propose RelGuard, the first differentially private framework for GraphRAG, which provides formal privacy guarantees by mitigating connectivity leakage and protecting relations and their semantic meaning through relation removal. We formalize relation differential privacy (relation-DP), which bounds the distinguishability of mechanism outputs according to the semantic distance between alternative relations while accounting for relation removal. Building on this definition, we develop a four-step defense mechanism that sanitizes source documents before GraphRAG indexing by perturbing relation predicates or removing their associated connectivity using the exponential mechanism. Experiments across three GraphRAG systems and two targeted reconstruction attacks demonstrate that RelGuard provides stronger protection against reconstruction attacks than four existing defense mechanisms while better preserving downstream utility.
Reza Rasti, Armin Biglari, Sobhan Mohammadi Hafshejani, Alireza Dehghani, Michael J. Allingham, Sina Farsiu
原文摘要
Diabetic macular edema (DME) is a leading cause of severe vision loss among working-age adults. Although intravitreal anti-vascular endothelial growth factor (anti-VEGF) injections currently serve as the first-line therapy for DME, a notable subset of patients fails to respond despite the substantial cost and clinical burden, suggesting they may benefit more from alternative interventions. Automated, noninvasive pre-treatment response assessment could therefore spare non-responders from the current trial-and-error treatment paradigm. In this study, we propose a novel conditional adversarial diffusion framework that synthesizes post-treatment optical coherence tomography (OCT) images from baseline scans, enabling the derivation of each patient's initial response to anti-VEGF therapy. The model was developed with subject-level 5-fold cross-validation on a US cohort (n = 127) and externally validated on an independent Iranian cohort (n = 44). Benchmarked against fourteen baseline and state-of-the-art translation models, the proposed model achieved the best treatment-response prediction performance among the methods evaluated for that task on both cohorts, with balanced accuracies of 85.7% and 93.0% on the US cohort and 85.2% and 95.3% on the external Iranian cohort at the 10% and 20% central subfield thickness reduction thresholds, respectively. Ultimately, this framework offers a decision-support tool to help ophthalmologists anticipate initial anti-VEGF outcomes in DME prior to treatment initiation, toward personalized therapy selection. The source code will be made publicly available upon acceptance.
A regular expression specifies a set of strings formed by single characters combined with concatenation, union, and Kleene star operators. Given a regular expression $R$ and a string $Q$, the classic regular expression matching problem is to decide whether $Q$ matches any string specified by $R$. Let $m$ and $n$ be the lengths of $R$ and $Q$, respectively. Our main result is a new algorithm that, given a parameter $t \geq m$, solves the problem in \[ O\left(\frac{nm\,\lg\lg(t/m)}{\lg^2(t/m)}+t+n\right) \] time and $O(t)$ words of space. In particular, for any fixed $0<\eps<1$, when $m\leq n^{1-\eps}$, we solve the problem in $O(nm \lg \lg n/\lg^2 n + m + n)$ time and $O(n)$ space. Compared to the previous best result [Bille and Thorup, ICALP 2009], this improves the leading term by a factor $\sqrt{\lg n}/\lg\lg n$. The key new technique is a two-dimensional tabulation specialized to the regular expression.
Conventional online surveys provide limited personalization, often resulting in low engagement and superficial responses. Although AI survey chatbots improve convenience, most are still reactive: they rely on fixed dialogue trees or static prompt templates and therefore cannot adapt within a session to fit individual users, which leads to generic follow-ups and weak response quality. We address these limitations with AURA (Adaptive Understanding through Reinforcement Learning for Assessment), a reinforcement learning framework for AI-driven adaptive conversational surveys. AURA quantifies response quality using a four-dimensional LSDE metric (Length, Self-disclosure, Emotion, and Specificity) and selects follow-up question types via an epsilon-greedy policy that updates the expected quality gain within each session. Initialized with priors extracted from 96 prior campus-climate conversations (467 total chatbot-user exchanges), the system balances exploration and exploitation across 10-15 dialogue exchanges, dynamically adapting to individual participants in real time. In controlled evaluations, AURA achieved a +0.076 mean gain in response quality and a statistically significant improvement over non-adaptive baselines (p=0.044, d=0.66), driven by a 63% reduction in specification prompts and a 10x increase in validation behavior. These results demonstrate that reinforcement learning can give survey chatbots improved adaptivity, transforming static questionnaires into interactive, self-improving assessment systems.
Given an arbitrary function $f:[N]\to[M]$, the function inversion problem asks to preprocess $f$ and to be able to, when given some $y\in [M]$, decide whether there exists some $x\in [N]$ such that $f(x)=y$ and if so, to report such an $x$. Recently, Fiat-Naor's solution to this problem has seen significant traction in theoretical computer science, and allowed for several conjectures related to meta-complexity and the 3SUM problem to be broken. In this paper, we consider the natural extension of the function inversion problem that we call preimage reporting. Here, instead of having to report only one $x\in [N]$ such that $f(x)=y$, the whole set $f^{-1}(y)$ must be found. We present a new general method to convert any arbitrary function inversion scheme into an efficient corresponding preimage reporting scheme. In particular, if $Δ$ is the maximal indegree of $f$, we show that we can perform preimage reporting using space $S$ in time $T$ for any $T$ and $S$ satisfying $TS^2=N^2 Δ$, matching a variant of a tradeoff obtained by Fiat and Naor for function inversion. This last result implies an improvement over the corresponding function inversion scheme itself, which constitutes the first function inversion scheme whose query time is inversely proportional to $|f^{-1}(y)|$. We derive new simple solutions for the $k$SUM-indexing with reporting and Jumbled indexing with reporting problems.
Prompt learning adapts vision-language models such as CLIP by optimizing a short learnable prompt. The cross-entropy objective pulls this prompt toward base-class specialization, whereas generalization to unseen classes benefits from staying close to CLIP's zero-shot feature space. Because both objectives act on the same prompt, their gradients conflict at convergence, and in our experiments no single-prompt loss design clearly improves on the cosine baseline. We propose Decoupled Mode Connectivity (DMC), which gives each objective its own prompt. Starting from a base-trained prompt, DMC trains a specialization prompt and a generalization prompt, and merges their text features with a fixed coefficient at test time, at the inference cost of a single prompt. A linear mode connectivity (LMC) corridor in text-feature space keeps the classification loss low for every classifier interpolated between the two endpoints, and a Visual Anchor KL term preserves CLIP's zero-shot class-similarity distribution of each image during specialization. We further argue that such regularizers should satisfy class-permutation invariance (CPI), and show via Fano's inequality, in an idealized model, that violating CPI lower-bounds the drop in novel accuracy by a term proportional to the mutual information between the prompt and the base-class labels. Across 11 datasets, DMC improves the average harmonic mean of base and novel accuracy over MergeTune for both CoOp and KgCoOp, and its interpolation curves pass beyond the single-prompt operating point, shifting the base-novel frontier rather than trading along it. Code is available at https://github.com/IemProg/DMC.
Ritik K. Badiya, Siladittya Manna, Ashmit Sinha, M Yashwanth, Anirban Chakraborty
原文摘要
Parameter-efficient fine-tuning (PEFT) has been increasingly adopted to fine-tune pretrained models with low computational cost and is therefore also preferred in the federated learning paradigm to reduce computational and communication overhead. In this work, we investigate the sparsely explored problem of the risk of private data leakage in LoRA-adapted Vision Transformers (ViTs), where low-rank updates weaken the gradient signal available to a server with access to individual client updates. We show that quality of reconstruction is limited by two coupled ambiguities: reduced information in LoRA gradients and missing spatial cues from frozen positional embeddings. To address this, we propose a spatially anchored generative inversion framework that incorporates server-prescribed trainable positional embeddings into ViT blocks during fine-tuning at clients. The resulting positional gradients act as layer-wise spatial anchors, providing additional positional cues, while the generative prior keeps reconstructions on a realistic image manifold, reducing duplicate-data and patch-permutation artifacts, {which produce similar gradients as those produced when the target client images are used}. Adding our proposed spatial anchoring to GIAS improves PSNR by 10.14 dB, SSIM by 0.45, and LPIPS by 0.098 on CIFAR-10, as well as PSNR gains of 7.83 and 11.74 dB on high-resolution CelebA and CARS datasets, respectively, in the FedSGD setting. Such gains in reconstructed image quality metrics are also maintained at higher batch sizes and in FedAvg settings. These results indicate that LoRA-based PEFT does not, by itself, prevent data leakage in FL, and carefully incorporated positional embeddings can help in exposing private information.
Multi-output Gaussian process (MOGP) regression allows modelling dependencies among multiple correlated response variables. Similarly to standard Gaussian processes (GPs), MOGPs are sensitive to model misspecification and outliers, which can distort predictions within individual outputs. In the multi-output case, this situation can be further exacerbated as anomalous observations in one response can adversely influence predictions for other correlated outputs. To handle this situation, we propose R-MOGP, an MOGP that is provably robust to outliers in any individual response. Our approach preserves conjugacy, achieving robustness without sacrificing computational efficiency, and integrates naturally with sparse variational inference for scalability. Empirically, we show that R-MOGP achieves performance comparable to existing state-of-the-art robust MOGPs at a fraction of their computational costs.
Ghjulia Sialelli, Robin Young, Cesar Aybar, Yuchang Jiang, Linus Scheibenreif, Damien Robert, Clemens Mosig, Kristof Van Tricht 等 15 位作者
原文摘要
The past decade has seen an explosion of large-scale geospatial products derived from Earth observation (EO) data using machine learning (ML). Driven by expanding model capacity and compute availability, the domain has reached a level of operational maturity where practitioners can generate high-resolution global products spanning thematic areas as diverse as human settlement, vegetation structure, and natural hazards. Deep learning has expanded the paradigm beyond locally calibrated models toward singular, globally applicable models trained on heterogeneous data, attracting a broad and growing community of practitioners from both the EO and ML sides. Yet, the resulting workflow introduces challenges that fall between established EO practice and mainstream ML: every stage presents tightly coupled technical challenges that recur across applications, and much of the practical knowledge needed to navigate them remains siloed within individual application domains. This guide brings together map-makers across the EO and ML communities, distilling their common practices, recurring pitfalls, and recommended approaches into a single end-to-end account, from raw satellite data to a published map product. Crucially, it treats the pipeline as an integrated workflow rather than a sequence of independent steps. We focus on wall-to-wall raster maps at moderate resolution (10-30m) and continental to global extent, produced with ML from open satellite archives, mainly Sentinel-1, Sentinel-2 and Landsat. The article is written for a deliberately broad audience, from those just entering the EO domain to ML experts building ML-ready EO datasets or generating large-scale maps.
Jun Hatayama, Daichi Mukunoki, Yusuke Onishi, Atsushi Sekimoto, Toshinori Kouchi
原文摘要
Scientific computing often requires FP64 arithmetic, although FP64 performance is limited on NVIDIA GPUs. Double-float (DF) emulation, which represents a value using two FP32 values with about 48 bits of significand, outperforms native FP64 where FP64 units are limited. On the other hand, mixed-precision strategies offload to FP32 the operations that do not determine accuracy. In both approaches, the native FP64 units remain idle during FP32 instruction execution. We investigated whether this idle resource can be recovered on A100, GH200, GB200, and RTX PRO 6000 Blackwell Max-Q. A microbenchmark shows that co-resident FP32 and native-FP64 kernels exceed the throughput of the faster kernel alone on A100 but not on GH200 and GB200; on RTX 6000 Blackwell, they do so against FFMA but, against DF emulation, only when the issue density of the DF kernel is reduced. We then tested whether this gain carries over to computing the WENO-Z reconstruction with the nonlinear weights in FP32 and the candidate polynomials in native FP64 (W32P64). We observe substantial overlap on A100, whereas GH200, GB200, and RTX 6000 Blackwell show little overlap. Importantly, the fastest placement does not always preserve the design accuracy of WENO-Z: for a one-dimensional entropy wave problem, W32P64 fails to achieve the design order of accuracy for WENO7 and WENO9. In contrast, DF emulation of the entire reconstruction comes close to it on the grids tested for WENO5, WENO7, and WENO9 while delivering a $2.51-3.94 \times$ speedup over FP64 on RTX 6000 Blackwell. For WENO5 on the two-dimensional Euler vortex problem the W32P64 error at $n=256$ is $1.26 \times$ that of FP64 and 1/104 of that of the all-FP32 reconstruction.
Mutation testing for deep learning (DL) faces a cost-realism trade-off. The pre-training approach produces realistic but computationally expensive mutants, whereas existing post-training operators modify trained models efficiently but produce mutants that less closely resemble real faults. This study asks whether the trade-off is necessary: can the weight-space effect of a fault-informed pre-training mutation operator be predicted from a trained model, without retraining it? We answer this question with TensorMutation, which represents a trained model as a neural graph and uses a mutation-conditioned graph neural network (GNN) to predict parameter changes. The GNN is trained offline and reused across unseen models so that TensorMutation can generate executable mutants without retraining. We construct a corpus of 25,252 original-mutant pairs from 71 multilayer perceptron (MLP) classifiers using 11 fault-informed pre-training operators, including 25 newly curated subjects. TensorMutation partially reproduces the behavioral effects of pre-training mutations on nine held-out subjects, and direct weight-space supervision significantly improves weight-space fidelity. On 12 real-fault subjects, TensorMutation achieves the highest median behavioral similarity among the evaluated approaches. The improvement over the post-training tool DeepMutation++ is statistically significant (Holm-adjusted p=0.018), whereas the differences from the pre-training tool DeepCrime and the norm-matched random control are not significant. These results show that fault-informed pre-training mutations leave learnable structure in weight space that transfers to unseen models.