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

+1022新增论文
+813其中 AI 相关
+85已有论文修订
+95新增 AI 导读

arXiv 最近完成扫描:2026-11-17 04:58 · 全站最近入库:2026-11-17 13:07各信源进度 →

统计口径与数据新鲜度

统计 2026-11-16 16:04 — 2026-11-17 16:04(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:2026-11-17 04:58最近发现新增版本:2026-11-17 13:07库内论文最新更新时间:2026-11-16 17:16

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。

24,849 已收录论文20,754 AI 相关7,213 有修订112 HF 社区发现

浏览论文库

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 24,849 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-17 04:57(北京时间)
版本更新 · v3arXiv:2507.09608

prNet: Data-Driven Phase Retrieval via Stochastic Refinement

Mehmet Onurcan Kaya, Figen S. Oktem

原文摘要

Phase retrieval is an ill-posed inverse problem in which existing methods often struggle to balance reconstruction fidelity and perceptual quality. We propose a novel Langevin-based framework for phase retrieval that bridges model-driven and data-driven approaches by combining stochastic posterior sampling, learned denoising, and model-based updates to better navigate the perception-distortion tradeoff. The framework consists of three progressively more advanced variants. The first, prNet-Small, provides a theoretically grounded and lightweight pipeline that integrates Langevin dynamics with learned noise and denoising processes and alternating projection-based updates. Building upon this foundation, prNet-Large exploits parallel sampling of diverse reconstructions and aggregates them to approximate the MMSE solution, significantly improving distortion metrics. Finally, prNet-Large-Adversarial enhances perceptual quality by replacing simple averaging with an adversarially trained aggregation network while preserving reconstruction fidelity. In addition, we incorporate test time augmentation to further improve the performance. Extensive experiments demonstrate that the proposed framework consistently outperforms classical, deep learning-based, and diffusion-based baselines across multiple benchmarks in both reconstruction fidelity and perceptual quality. The source code and trained models are available at https://github.com/METU-SPACE-Lab/prNet-for-Phase-Retrieval

eess.IV · cs.CV

版本更新 · v2arXiv:2607.19120

Latent Riemannian Flow Matching for Geometry-Grounded 3D Foundation Models

Lisa Weijler, Irene Ballester, Guofeng Mei, Tolga Birdal, Pedro Hermosilla

原文摘要

Geometric foundation models such as the Visual Geometry Grounded Transformer (VGGT) provide strong 3D priors from unposed images, but operate deterministically and cannot generate plausible geometry beyond what the observations directly support. We bridge geometric foundation models and generative 3D modeling by performing flow matching directly in their latent space, leveraging learned 3D priors without committing to explicit representations such as Gaussians, meshes, or video-VAE latents. This requires respecting the latent geometry: LayerNorm places the decoder inputs on a product of zero-mean hyperspheres, and treating this space as unconstrained Euclidean substantially degrades flow matching. We therefore formulate Riemannian flow matching on this product manifold, conditioned on unposed context views and a target camera pose, keeping generated tokens within the geometry expected by the frozen decoder. Our main experiments use VGGT, while additional experiments with Depth Anything 3 show that the same formulation transfers to a different independently pretrained 3D backbone without modification. On RealEstate10K, ScanNet++, and ETH3D, our method achieves strong results against recent scene-generation baselines in both novel-view appearance and aggregated 3D geometry, establishing latent-space flow matching on geometric foundation models as a viable paradigm for 3D generation.

cs.CV

新论文 · v1arXiv:2611.08775

On Lipschitz Explosion in Deep Neural Networks with Normalization: Consequences for Optimization and Adversarial Robustness

Ashkan Soleymani, Reyhaneh Hosseinpourkhoshkbari, Hadi Daneshmand, Patrick Jaillet

原文摘要

Can a network look stable at every layer and still be exponentially unstable as a whole? The Lipschitz constant of a neural network is a basic stability quantity appearing in optimization guarantees, generalization bounds, and robustness certificates. Deep networks naturally admit two such notions, input Lipschitzness, measuring sensitivity to data perturbations, and parameter Lipschitzness, measuring sensitivity to weight perturbations. We prove that, for deep networks with normalization layers, with batch normalization as the canonical example, both Lipschitz constants can grow exponentially with depth and hence with parameter dimension, even under strong per-layer norm control such as $\|W_\ell\|_2\le 1$, and already for linear activations. Thus, a mechanism widely used to stabilize training can create severe worst-case instabilities that are invisible locally, i.e., from layerwise norm bounds alone. Parameter-Lipschitz explosion turns Lipschitz-dependent nonsmooth optimization guarantees into exponential-in-dimension bounds for deep normalized networks, while input-Lipschitz explosion makes worst-case Lipschitz-based generalization and robustness certificates vacuous. The latter also yields a concrete rank-separation mechanism for adversarial vulnerability. The theory predicts that perturbations introducing new singular directions should be amplified much more strongly than perturbations at the same noise level that remain within the input's original singular subspace. Experiments on MNIST, Fashion-MNIST, and CIFAR-10 support this prediction, showing that rank-creating perturbations cause substantially sharper drops in accuracy, confidence, and margins than same-subspace perturbations.

cs.LG · cs.AI · math.OC · stat.ML

新论文 · v1arXiv:2611.08774

Gain-Dependent Entropy Characterization of a Zener Avalanche Noise TRNG via Dual NIST Standards: Toward Low-Cost Hardware Entropy for ICS Deployments

Rishabh Das

原文摘要

True random number generators based on Zener avalanche noise offer a low-cost entropy source for industrial control system and IoT edge devices, yet the relationship between amplifier gain, coupling capacitance, and entropy quality remains uncharacterized under current NIST standards. This paper presents a comprehensive entropy assessment of an op-amp-based Zener avalanche noise generator across five gain configurations and two coupling capacitor values, evaluated against the NIST SP 800-22 statistical test suite and the SP~800-90B entropy source assessment using 1.44 billion raw bits per configuration. We identify a three-phase autocorrelation structure, attributed to comparator saturation-recovery dynamics, consistent across all gain settings. A direct-current substitution experiment, replacing the Zener diode with a stable voltage source at the same bias point, reduces switching activity by approximately 97%, confirming that Zener avalanche noise, rather than amplifier self-excitation, is the necessary entropy source. The coupling capacitor is further shown to govern comparator overdrive and entropy quality independent of this source mechanism, with the smaller capacitance improving entropy, throughput, and bias. The combined results identify an optimal gain configuration achieving a perfect statistical test score, NIST-compliant min-entropy, and peak entropy throughput at a bill-of-materials cost under five dollars, offering practical design guidance for low-cost entropy sources in industrial and embedded deployments.

cs.CR · cs.AR

版本更新 · v5arXiv:2603.14955

Convex algebras on an interval with semicontinuous monotone operations

Ana Sokolova, Harald Woracek

原文摘要

In a recent work of Matteo Mio on compact quantitative equational theories (here compact means that all its consequences are derivable by means of finite proofs) convex algebras on the carrier set [0,1] whose operations are monotone and satisfy certain semicontinuity properties occurred. We fully classify those algebraic structures by giving an explicit construction of all possible convex operations on [0,1] possessing the mentioned properties. Our result thus describes exactly the range of theories to which Mio's theorem applies.

cs.LO · math.LO

版本更新 · v2arXiv:2601.11836

OmniLink: Specification-Based Trace Validation of Concurrent Systems with Substep Linearizability

Finn Hackett, Evan Wrench, Peter Macko, A. Jesse Jiryu Davis, Yuanhao Wei, Ivan Beschastnikh

原文摘要

Concurrent systems are notoriously difficult to validate. We present OmniLink, a methodology for validating traces of concurrent implementations against high-level specifications in TLA+. Unlike prior TLA+ based approaches, OmniLink treats system events as black boxes with a timebox in which they occurred and a meaning in TLA+, solving for a logical total order of actions. Unlike prior approaches based on linearizability checking, OmniLink implements substep linearizability semantics, relies on a flexible specification language, accelerates checking with user-customizable path pruning specifications, and can detect violations due to weak memory effects between operations. Our evaluation validates WiredTiger, a state-of-the-art industrial database storage layer; Balanced Augmented Tree (BAT), a state-of-the art lock-free data structure from the research community; and, ConcurrentQueue, a popular lock-free queue featuring aggressive performance optimizations. OmniLink detects injected and previously fixed bugs in the systems under test. It also uncovers a previously unknown bug in BAT that has been confirmed by its authors.

cs.SE

版本更新 · v2arXiv:2611.05797

DARE: Difficulty Analysis for Repair Enhancement

Jikwang Lee, Seungwook Jung, Jaechang Nam

原文摘要

LLM-based coding agents are primarily evaluated by their solve rates. However, this metric does not reveal how difficult the solved tasks are or account for differences in task difficulty across benchmarks. Determining the difficulty of individual tasks by repeatedly running multiple agents is costly, while a predicted difficulty score alone provides limited guidance on what support an agent needs. To address these challenges, we propose DARE (Difficulty Analysis for Repair Enhancement), which integrates repair-task difficulty measurement, prediction, feature-group contribution analysis, and task-specific repair support. We define continuous task difficulty as the observed failure rate across multiple coding agents. The DARE difficulty model predicts this difficulty using 151 static features extracted from benchmark artifacts. The framework also analyzes the contributions of seven Repair-Task Feature Groups to identify the characteristics most salient to each task and uses this information to select task-specific agent support. We evaluate DARE on four benchmark partitions comprising 2,594 tasks. The difficulty model's aligned predictions achieve Spearman's $ρ$ = .476-.621 on held-out partitions, and the most salient Feature Groups vary across tasks and benchmarks. Furthermore, coding-agent experiments show that matching support to each task's salient Feature Group raises repair success from 54.7% to 79.6% on nine high-difficulty tasks. These findings demonstrate that difficulty prediction and interpretation can lead to practical improvements in automated program repair (APR), while also providing a basis for difficulty-stratified agent evaluation, cross-benchmark comparisons, and identifying areas of repair capability that warrant further improvement.

cs.SE · cs.AI

版本更新 · v2arXiv:2607.18659

Broken Gates: Re-evaluating Web Bot Defenses in the Age of LLM Agents

Behzad Ousat, Nikita Turkmen, Lalchandra Rampersaud, Dillan Bailey, Amin Kharraz

原文摘要

The rapid emergence of web automation across commercial crawlers, solver services, and LLM-based browser agents has created new challenges for defenses that must distinguish legitimate automated activity from scalable abuse. In response, modern bot defenses increasingly replace explicit challenges with risk-based decisions over browser, session, identity, and behavioral signals. This paper explores how current automation stacks can bypass these defenses end-to-end, evaluating seven commercial solver services and eight LLM-based browser agents against four widely deployed defenses. We find that solver services largely defeat challenge-based defenses with a 100% success rate, but collapse to an average 23% success rate on reCaptcha v3. Browser agents encounter a different bottleneck: the browser and account state presented to the verifier. Controlled experiments across browser configurations show that an established consumer account and persistent Chrome state strongly affect reCaptcha v2/v3 acceptance in our deployments, whereas Cloudflare Turnstile exhibits substantially greater sensitivity to the agent's browser control path. Reconstructing the established browser environment enables previously rejected agents to pass without model retraining. Once established, this trusted environment transfers across the agents and defenses in our testbed, adding no measurable defense-specific marginal cost to the surrounding workflow. These results expose an inherent security tension in modern non-interactive defenses: to eliminate friction, the signals they trust must be accessible to everyday human users, yet this accessibility inherently guarantees that adversaries can acquire, reproduce, and transfer the same identity signals to scale automated abuse.

cs.CR · cs.AI · cs.MA

版本更新 · v3arXiv:2609.04566

Optimizing Credential Blast Radius Through Trust Boundaries and Delegation Under Post-Quantum Authentication Costs

Pauli Taipale, Harri Lainio

原文摘要

Partitioning interacting services into independently rooted trust domains limits issuer-compromise reach while increasing calls across trust boundaries. Post-quantum replacements for public-key authentication and key-establishment mechanisms can increase crossing latency on constrained or lossy paths. We formulate the joint selection of trust domains and credential-derivation structures under policy and latency constraints, linking separate service-interaction and credential-derivation graphs through domain assignment. Credential blast radius measures weighted service impact after compromise. A linear upper bound supports optimization, while a joint event model gives exact expected impact. We identify when risk from issuers trusted across domains can be incorporated into this linear score, avoiding separate issuer-propagation calculations for each candidate. Although the general problem is NP-hard, we identify restricted cases that can be solved efficiently and exactly. Joint optimization yields lower blast radius than choosing boundaries first in 195 of 230 exhaustive synthetic comparisons, especially under chained delegation. A trace-derived replay used measured post-quantum costs, synthetic risk inputs, a fixed derivation family, and one to six trust domains. Under independent compromise events, mean expected impact was up to 36% lower than with one domain within the latency budget. The framework turns risk assumptions and measured crossing costs into candidate trust-domain and credential-derivation designs.

cs.CR

版本更新 · v2arXiv:2605.07459

On the Complexity of Discounted Robust MDPs with $L_p$ Uncertainty Sets

Ali Asadi, Krishnendu Chatterjee, Alipasha Montaseri, Ali Shafiee

原文摘要

A basic model in sequential decision making is the Markov decision process (MDP), which is extended to Robust MDPs (RMDPs) by allowing uncertainty in transition probabilities and optimizing against the worst-case transition probabilities from the uncertainty sets. The class of $(s, a)$-rectangular RMDPs with $L_p$ uncertainty sets provides a flexible and expressive model for such problems. We study this class of RMDPs with a discounted-sum cost criterion and a constant discount factor. The existence of an efficient algorithm for this class is a fundamental theoretical question in optimization and sequential decision making. Previous results only establish a strongly polynomial-time algorithm for $L_\infty$ uncertainty sets. In this work, our main results are as follows: (a)~we show that for any compact uncertainty set, the policy iteration algorithm for RMDPs is strongly polynomial with oracle access to solutions of Robust Markov chains (RMCs); (b)~we present strongly polynomial-time bounds on the policy iteration algorithm for RMCs with $L_1$ and $L_\infty$ uncertainty sets; and (c)~we establish hardness results for RMCs with $L_p$ uncertainty sets for integer $p$ satisfying $1<p<\infty$. Finally, motivated by our theoretical bounds, we present experimental results showing how fast policy iteration converges for RMDPs with $L_1$ and $L_\infty$ uncertainty sets.

cs.CC

版本更新 · v2arXiv:2610.35362

Computation-Friendly Compression of Matrices: Speeding up Linear and Boolean Algebra

Gonzalo Navarro

原文摘要

Large sparse matrices---adjacency matrices of web and social graphs, feature matrices of machine-learning pipelines, the relations of graph databases---are routinely compressed to save space. Classical compressors, however, force the matrix to be decompressed before it can be used, so the space savings often yield a computational overhead. A recent line of work studies \emph{computation-friendly} compression: formats that are not only small, but let us carry out matrix--vector products, and in some cases matrix--matrix products, sums and closures, directly on the compressed representation {\em in time proportional to that compressed size}. This survey organizes that literature around the kind of redundancy each format exploits: spatial clustering of the nonzeros ($k^2$-trees and their variants), similarity between rows (Boldi--Vigna-style referencing and set-difference compression), repeated substrings (grammar compression), and dense bipartite substructures (bicliques). For each family we describe the representation, the algorithms, the theoretical guarantees and the empirical performance. Further, we show that the last three families are instances of a single general scheme, which we call a {\em sparse straight-line factorization} of the matrix, whose forward and backward substitutions evaluate right and left products in time proportional to the size of the representation, and that the product of two such factorizations is again one, which gives matrix--matrix products with a compressed output. We also relate the model to linear circuits, give simple recompression steps, and study the size of the smallest factorization: worst-case bounds, separations from each family, and parallelism. We close with a cross-family comparison and a list of open problems.

cs.DS

版本更新 · v3arXiv:2605.16757

NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning

Haoran Lu, Luyang Fang, Wenxuan Zhong, Ping Ma

原文摘要

Multi-agent language systems are often built as hand-designed workflows, where agents are assigned semantic roles and communication protocols are specified in advance. We propose NeuroMAS, a framework that treats a multi-agent language system as a trainable and scalable neural-network-like architecture with hidden textual intermediate signals. In NeuroMAS, language-agent nodes are role-free but structure-aware: the topology determines how information can flow, while reward-based training determines how nodes communicate, specialize, and coordinate. This formulation shifts multi-agent design from workflow engineering toward architecture design, where depth, width, connectivity, and growth protocol become scalable sources of capability. We provide a theoretical perspective showing why such modular text-valued computation can be parameter-efficient when tasks admit stable decompositions. Experiments show that NeuroMAS improves over both inference-time collaboration methods and trained baselines, and that its gains are not explained by trainable parameter count alone. We further find that organizational scaling is path-dependent: larger systems can fail when trained from scratch, but become stronger when grown progressively from smaller trained systems. These results suggest that learned multi-agent organization is a promising scaling axis for language-model reasoning and generation.

cs.AI · cs.MA · stat.ME · stat.ML

新论文 · v1arXiv:2611.08772

StarLoc: A Practical Approach to Encoding Geographic Locators in IPv6 for Starlink

Vanya Singla, Mostafa Abdollahi, Jinwei Zhao, Jianping Pan

原文摘要

LEO satellite networks such as Starlink provide Internet connectivity across continents, oceans, and airspace. IP geolocation providers map Starlink's IPv6 prefixes to locations mainly through its self-published GeoFeed. However, users may be far away from their assigned Points of Presence (PoPs), one label covers an entire prefix, and labels and PoP mappings can change while addresses stay the same. We characterize Starlink's IPv6 addressing using 925 historical GeoFeed snapshots (2024--2026), one year of PoP mappings, and over 7.9 million observed IPv6 addresses. Our findings reveal that customer prefixes lack clear geographic structure, prefix labels lie a median of 540km from their PoPs, and relabeling a single /40 can affect the reported locations of up to 65,536 possible /56 delegations. To address these challenges, we propose StarLoc, a geographic Locator--Identifier separation scheme that embeds terminal location into IPv6 addresses using a resolution-adaptive H3 Locator while preserving existing Starlink prefixes, /56 delegations, and stable Endpoint Identifiers. To balance accuracy and privacy, StarLoc reveals only cells with at least a certain number of terminals, and operators can cap resolution in sensitive regions or withhold location entirely. Unlike topology-based schemes, StarLoc encodes physical location and lets Locators follow moving users.

cs.NI

新论文 · v1arXiv:2611.08771

Federated training performs as well as centralized for 3D Latent Diffusion for Diagnosis-Conditioned Brain MRI Synthesis

Abhijith Shaji, Paul Kurian, Greg Ver Steeg, Paul M. Thompson, Jose-Luis Ambite

原文摘要

Generative models of brain MRI are useful to understand brain conditions, and to enrich scarce and imbalanced neuroimaging data. However, training these models requires large, multi-site training sets, which is challenging. Federated learning (FL) allows for training deep learning models without sharing data. We show that training a 3D conditional latent model federatedly (where data is kept private at each site) performs similarly to centralized training (where all the data have been copied to a single site). We evaluate three metrics: conditional fidelity (how well generated MRIs perform in classification), diversity (how the generated images match the real data distribution, using MultiScale-SSIM), and realism (how close generated images are to real images, using Fréchet Inception Distance). Our results show that federated training is a viable route to develop multi-site generative models of brain MRI.

cs.CV · cs.AI

新论文 · v1arXiv:2611.10051

All roads lead to Markov kernel networks: A representation theorem for discretization-stable learning

S David Mis, Trevor Teolis, Ivan Dokmanić, Matti Lassas, Maarten V. de Hoop

原文摘要

Operator learning seeks to approximate function-to-function maps in a way that remains stable when the input discretization changes, a property sometimes called "discretization invariance". We consider a quantitative refinement of this idea, called $\textit{uniform discretization stability}$, which builds on a common representation of input functions and their discretizations as probability measures and provides a Hölder bound relating output changes to the Wasserstein distance between these measures. Our main result is a representation theorem showing all uniformly discretization stable operators can be written exactly as a $\textit{stable Markov kernel network}$---an abstract architecture that is a joint generalization of transformers, Fourier neural operators (FNO), Flowers, and other successful learned operators. Our representation theorem characterizes a design space for discretization-stable operator learning; designing new architectures becomes a question of which parts of a Markov kernel network to parameterize, and how. We further show that the read kernels of these networks can be realized as solutions of regularized variational problems, with particular choices of costs and regularizers recovering the kernels of transformers, FNO, Flowers, their hybrids, and beyond. This variational formulation provides a systematic approach to architecture design, together with sufficient conditions for uniform discretization stability.

cs.LG

新论文 · v1arXiv:2611.10050

Selected-Risk Calibration for Feasibility at Returned Decisions

Raghav Thind, Youran Sun, Ling Liang, Haizhao Yang

原文摘要

When an optimization problem uses a surrogate constraint, a decision that is feasible for the surrogate need not be feasible for the true constraint. We propose Selected-Risk Calibration (SRC), a framework that calibrates a learned constraint against the optimizer's returned decisions for objectives drawn from a given distribution. SRC guarantees that, with probability at least $1-δ$, every margin it certifies on held-out objectives has selected risk, the probability that the returned decision is infeasible, of at most $γ$. We also show that the random-input coverage guarantee of split conformal prediction does not bound the selected risk, and that likelihood-ratio reweighting for feedback covariate shift does not apply when the learned constraint is active at the returned decision. Our experiments show that a conformal margin covering about $90\%$ of random inputs covered as few as $17\%$ of the returned decisions, and every certified margin had a violation rate below its target on unseen objectives. Moreover, at equal cost, SRC's correction had fewer or equal infeasible solutions to the baselines in 57 of 62 comparisons. The SRC framework applies to any setting in which a surrogate model chooses a decision whose feasibility can be verified for a sample of instances.

cs.LG · stat.ML

新论文 · v1arXiv:2611.08770

ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation

Guangjian Dong

原文摘要

Prefill--decode disaggregation moves a prompt's key--value (KV) cache between workers before the first token can be generated. Reducing this handoff to either cache compression or recomputation misses a useful choice: which layers to transmit, at what precision, and which layers to execute again on the decode worker. ReKV transfers a quantized prefix KV cache and one BF16 boundary state, then reconstructs a contiguous suffix while the prefix is in flight. In a model-shaped synthetic benchmark on cross-node H100s, the Q8 configuration meeting separately evaluated quality criteria reduces 8K handoff latency by $2.36\times$, $2.84\times$, and $3.65\times$ over BF16 transfer for Qwen2.5-7B, Qwen2.5-3B, and Qwen3-8B. It is also faster than the best measured transfer-only or recompute-only action at those points. A separate real-weight BF16 recomputation experiment reduces prefill-to-token-return time by $1.16$--$1.63\times$. The preferred action changes with prompt length and link bandwidth.

cs.DC

版本更新 · v2arXiv:2611.04085

Follow the Features: SAE Feature-Guided Data-Efficient Instruction-Tuning of Large Language Models

Ryan Swift, Konstantinos Psounis

原文摘要

Selecting small, effective instruction-tuning datasets requires identifying which aspects of the original data to retain. We introduce Follow the Features (FtF), a data-selection method that uses sparse autoencoder (SAE) features to characterize training data and guide subset construction. FtF ranks features using dataset-level activation statistics, scores samples by their expression of highly ranked features, and incorporates feature resampling to encourage diversity. The LLM-SAE pair used for selection is independent of the model being fine-tuned, enabling reuse across target models. We evaluate FtF on two model-dataset settings at a training budget of 100 samples with a fixed sequence length of 256 tokens. FtF achieves performance competitive with GRAPE and S2L, with relative rankings varying across settings. On Qwen 3 8B trained on Alpaca, FtF exceeds full-dataset training by 1.59 percentage points in the harmonic mean of an AlpacaEval-based win rate and Open LLM Leaderboard performance, although full-dataset training retains a higher Leaderboard score. These results support SAE feature statistics as a useful signal for data-efficient instruction tuning.

cs.LG

版本更新 · v2arXiv:2611.04081

RAPID: Increasing Efficiency of Internal Distillation via Randomized Methods

Ryan Swift, Konstantinos Psounis

原文摘要

Knowledge distillation is a technique used to transfer the expertise of large teacher models into small, efficient student models. Recent advances have integrated the distillation of internal activations, more directly training students to learn similar representations as the teacher. One of the core challenges of internal distillation is the dimensional incompatibility between the student and teacher internal representations. Existing methods add learnable matrices to project the larger teacher activations into the smaller representation space of the student. This is a notable point of inefficiency, as learning additional parameters for the projectors increases training costs, only for them to be discarded immediately after training. We introduce RAPID (RAndomized Projection for Internal Distillation) a novel distillation methodology which eliminates the wasteful projection learning by leveraging randomized methods. RAPID utilizes an efficient transform construction based on the Johnson-Lindenstrauss lemma to compress the teacher representations into the student embedding space with strong theoretical guarantees of structural preservation. RAPID also employs a combination of exponential moving average (EMA) and homoscedastic uncertainty weighting in a novel multi-objective function designed for internal distillation. Our experiments in post-training for reasoning demonstrate that RAPID increases training efficiency by as much as 21.1% while maintaining performance comparable to existing methods.

cs.LG

新论文 · v1arXiv:2611.08767

MEDALLION: Manifold Embedding Distillation and Attributions for Low-dimensional Latent InterpretatiON

Vicky Jung, Claire M. He, Irene Chang, Genevera I. Allen

原文摘要

Dimensionality reduction is widely used to visualize high-dimensional data and guide scientific discovery, yet interpreting which original features contribute to low-dimensional representations remains challenging, particularly for black-box methods such as t-SNE and UMAP. We introduce \textbf{MEDALLION} (Manifold Embedding Distillation and Attributions for Low-dimensional Latent InterpretatiON), a novel framework for interpreting low-dimensional embeddings through knowledge distillation and neural feature attribution. MEDALLION distills any given embedding with a neural encoder which we explain via a broad range of existing neural feature attribution methods. Through experiments on MNIST and applications in single-cell transcriptomics, we show MEDALLION provides informative feature attributions that cast additional insights into the underlying embedding and compare its performance with existing interpretability approaches.

cs.LG · stat.ML

版本更新 · v3arXiv:2610.23047

DIFFLET: A Fast Inference Engine for Diffusion Transformer Models on AWS Trainium

Bin Ma, Min Yu, Vijay Kodamalla, Johnny Liu, Da Li, Dong Li

原文摘要

On AWS Trainium2, every component of a diffusion model runs as a program compiled ahead of time for fixed shapes, while pipelines differ across models, prompt lengths across requests, and caching decisions across denoising steps. We present Difflet, a serving engine in which five model pipelines share one compilation backend, joint attention is split across cores by attention head, and a host controller decides at each denoising step whether to run the denoiser or reuse its earlier output. A small probe program computes the caching signal on the device and returns one number per step, and a resident model process keeps every program loaded across requests. Across five image and video model families, fused attention accelerates denoiser steps by 1.2-4.3x over the same framework using PyTorch/XLA attention, and at 32.8k visual tokens it is the only attention that compiles. On four of the five models, resident serving is 1.5-3.4x faster than standalone runs that reload weights from a warm file cache for each generation.

cs.DC · cs.LG

版本更新 · v2arXiv:2610.10854

KDFP: A first-principles approach to knowledge distillation in large language models

Ryan Swift, Konstantinos Psounis

原文摘要

Knowledge distillation is an established technique for improving the capabilities of small, efficient student models by training them with the representations of larger, more capable teacher models. Much of the recent work in the distillation of large language models (LLMs) has focused on distilling abilities learned during post-training, such as instruction following, chain-of-thought reasoning, and tool usage. This has left a large research gap in general knowledge distillation for LLMs, which is essential for developing efficient and private systems suitable for deployment on edge devices. We take a first-principles approach, evaluating previous lessons from prior works and conducting new explorations to develop a distillation methodology suitable for modern LLMs. We present KDFP, a novel methodology for white-box general knowledge distillation in LLMs. We demonstrate that KDFP outperforms existing methods by 1.6% $-$ 4.9% across 9 benchmarks while increasing training efficiency by up to 99.1% through ephemeral parameter reduction.

cs.LG

版本更新 · v3arXiv:2509.00031

End-to-End On-Device Quantization-Aware Training for LLMs at Inference Cost

Qitao Tan, Xiaoying Song, Ningxi Cheng, Guoming Li, Jun Liu, Lingzi Hong, Caiwen Ding, Jundong Li 等 14 位作者

原文摘要

Quantization reduces the deployment cost of large language models (LLMs). However, post-training quantization (PTQ) cannot fine-tune model parameters and often degrades accuracy in low-bit settings, while quantization-aware training (QAT) requires backpropagation and incurs high memory overhead. We propose ZeroQAT, a zeroth-order QAT framework supporting both weight and activation quantization. By using forward-only gradient estimation, ZeroQAT removes backpropagation and substantially reduces memory and computation while retaining end-to-end optimization benefits. We further introduce a lightweight variant of ZeroQAT for quantized fine-tuning, which freezes and pre-quantizes most parameters to further cut memory usage. Experiments show that ZeroQAT consistently outperforms representative PTQ and QAT baselines while requiring significantly less memory. For example, ZeroQAT enables fine-tuning of a 13B model at extremely low bit-widths (e.g., 2-4 bits) on a single 8GB GPU, and even allows fine-tuning a 6.7B model on a OnePlus 12 smartphone, demonstrating its practicality for end-to-end QAT on resource-limited edge devices.

cs.LG · cs.AI

新论文 · v1arXiv:2611.10049

LoRazor: Trimming Adapter Execution for Efficient Multi-LoRA Serving

Xiang Li, Pengcheng Wang, Azam Ikram, Saurabh Bagchi

原文摘要

LoRA offers a parameter-efficient adaptation strategy, supporting multi-user workloads via customized adapters on a shared backbone. However, serving concurrent requests with different adapters introduces substantial computation and memory-management overhead. We ask the research question: Do all requests need their adapters throughout the whole generation process? Our motivational study reveals that employing LoRA adapters at every decoding step is unnecessary, releasing them early has limited impact on accuracy while substantially accelerating generation. Although skipping adapters early offers theoretical workload reduction, translating these computational savings into concrete throughput gains requires tailored GPU kernels. We present LoRazor, a multi-LoRA serving framework that combines early adapter release with active-row aware GPU kernels. LoRazor determines when to permanently release a request's adapter and continue decoding with the backbone alone. By dynamically resizing LoRA kernel launches as adapters are released, LoRazor trims idle GPU blocks. Across two backbones and four tasks, LoRazor achieves up to 3.02x request throughput and 2.53x output-token throughput over vLLM multi-LoRA, with a mean relative task-accuracy decrease of 2.89%.

cs.AI

版本更新 · v3arXiv:2604.17120

Sloan's Analytical Gömböc: Strict Convexity Narrows the Admissible $β$ Range

Vincent Wesley Couey

原文摘要

Varkonyi and Domokos (2006) proved that convex homogeneous bodies with exactly one stable and one unstable equilibrium point exist. Sloan (2023) gave the first analytical parameterization, with radial function $R(θ,φ)$ having exactly two critical points on $S^2$. This is the v3 amendment-of-record of arXiv:2604.17120 (v3 corrects an attribution in the v2 title and abstract; no finding changes). v1 claimed Sloan's parameterization does not produce mono-monostatic bodies and reported a 13-member catalog of Fourier/radial extensions certified at ECS=1 via mesh-vertex drainage-basin analysis. Following correspondence with P. L. Varkonyi (BME), an analytical verification suite was built around the Varkonyi-Gauss identity. Finding 1: Sloan's parameterization does produce mono-monostatic bodies in a strictly-convex sub-regime ($β\lesssim 0.036$), where $K_{\min} > 0$ and the identity certifies ECS=1. v1 missed this because its mesh-vertex oracle over-counted on shallow COM-height landscapes. Sloan states that $β\le 0.17$ should suffice for convexity, but at $β=0.05$, inside that range, strict convexity is lost ($K_{\min}=-0.569$; $K<0$ over 4.01% of surface) and the identity's precondition fails. v1's "global surface information" mechanism is replaced by the strict-convexity precondition. Finding 2: Of v1's 13 catalog instances only Phase-1 ($β=0.023149$, $a_1=0.234433$, $k=1$) survives identity-based verification; the remaining twelve were per-$k$ optimizer extrema overshooting the strict-convex boundary. Probing the regime interior verifies further mono-monostatic bodies in $k=2$ and $k=3$ sub-families: the verified set is an open regime in $(β, a_1, k)$, not a discrete list. Finding 3: v1's ECS=1 readings for the 9 radial-family members reflected drainage-basin merging; the $r=0.9993$ gentleness-robustness correlation is retracted.

cs.CE · cs.CG

新论文 · v1arXiv:2611.08763

CrowdDirector: Runtime Orchestration of Crowd Simulation through Natural-Language Directives

Xiaohan Sun, Vuk Radovanovic, Merce Alvarez, Carol O'Sullivan, Yiorgos Chrysanthou, Panayiotis Charalambous

原文摘要

Natural language offers an intuitive way to direct a virtual crowd while a simulation is running. Handwritten rules execute efficiently but cannot respond flexibly to varied instructions, whereas asking a large language model (LLM) to control the crowd at every update introduces repeated latency, cost, and the risk of unsupported actions. We present CrowdDirector, a framework that uses an LLM only when the user adds or changes an instruction. The LLM translates the request into a structured instruction, and a scene validator checks whether the entities and actions it references exist in the current Unity scene. Accepted instructions are stored and remain continuously active as the simulation is running. At each simulation tick, a graph neural network uses the current scene state and active instruction to choose an available action and target for each agent, while Unity handles movement, interactions, and animation. The network is trained offline on simulated scenarios, with LLM guidance on how agents should respond to instructions. Our evaluations show the scene validator's feasibility decisions matched human judgements more often than the LLM's own assessment, and on a behavioural benchmark CrowdDirector follows user instructions more closely than handwritten rules while avoiding repeated LLM calls. It therefore provides a practical middle ground between fixed rules and continuous LLM control.

cs.GR · cs.AI · cs.MA

版本更新 · v2arXiv:2607.09993

Beyond Bayesian Nash: Learning Minimax-Regret Equilibria for Adversarial Team Games under Asymmetric Information

Naman Aggarwal, Jonathan P. How

原文摘要

Adversarial team games (ATGs) with asymmetric information, such as adversarial path-finding, goal search, and reachability games on graphs, require strategies that are robust to hidden opponent types, such as a hidden goal flag, and to deception. Under asymmetric information, deception is seen as strategic shifts in the type distribution such that the omniscient opponent can collude with Nature and condition its play on the observed type. Existing risk-neutral solution concepts, such as Bayesian Nash equilibrium (BNE), are sensitive to distribution shifts, while distributionally robust approaches provide guarantees only within a prescribed ambiguity set. To address these limitations, we introduce Probabilistically Robust Minimax-Regret Equilibrium (PR-MRE), a novel equilibrium concept that combines the distribution-free robustness of minimax-regret reasoning with probabilistic information from a nominal type distribution. PR-MRE minimizes worst-case regret over a high-confidence subset of the type space, providing protection against strategic redistribution of probability mass while avoiding the conservatism of fully distribution-free approaches. We show that, for finite normal-form Bayesian games, approximate PR-MRE computation can be formulated as a robust bilinear program and derive a tractable semidefinite relaxation. We then adapt this relaxation into a novel meta-solver within a robust double-oracle framework, PRMRE-PSRO, enabling population-based learning of approximate PR-MRE strategies via deep reinforcement learning best responses. Experiments on graph-structured adversarial team games demonstrate that PR-MRE discovers strategies with substantially improved worst-case performance across hidden types compared to risk-neutral Bayesian Nash equilibrium solutions, resulting in more robust behavior under strategic distribution shifts.

cs.GT · cs.AI · cs.LG · cs.MA

版本更新 · v2arXiv:2605.20079

Probability-Conserving Flow Guidance

Parsa Esmati, Junha Hyung, Amirhossein Dadashzadeh, Jaegul Choo, Majid Mirmehdi

原文摘要

Diffusion and flow-based generative models dominate visual synthesis, with guidance aligning samples to user input and improving perceptual quality. However, Classifier-Free Guidance (CFG) and extrapolation-based methods are heuristic linear combinations of velocities/scores that ignore the generative manifold geometry, breaking probability conservation and driving samples off the learned manifold under strong guidance. We analyse guidance through the continuity equation and show its effect decomposes into a divergence term and a score-parallel term defined invariantly across parameterizations. We prove the divergence term blows up structurally as sampling approaches the data manifold, motivating a time-dependent schedule alongside score-parallel attenuation. The resulting plug-and-play rule, Adaptive Manifold Guidance (AdaMaG), targets both terms at no additional inference cost. Finally, we show that several existing saturation and manifold preserving heuristics can be read as acting on one of these two terms. On image generation benchmarks, AdaMaG improves realism, reduces visible artefacts, and induces controlled desaturation under strong guidance.

cs.CV · cs.AI · cs.LG · eess.IV

新论文 · v1arXiv:2611.08756

Debate to Deceive: Poisoning LLM's Fact-Checking via Game-Theoretic Multi-Agent Debate

Liting Huang, Huizhi Liang, Yi Lei, Xingjun Ma, Varun Ojha

原文摘要

Knowledge attacks on Large Language Models, such as manipulating fact-checking judgments, pose significant risks to AI reliability. Existing approaches -- adversarial suffix/template injection and single-agent self-reflection -- lack robustness across diverse defenses and are prone to confirmation bias and costly trial-and-error, respectively. We propose Debate-to-Deceive (D2D), a multi-agent debate framework grounded in human debate structure and rhetorical style, enabling agents to generate adversarial passages that are semantically coherent, rhetorically effective, and resistant to early confirmation bias. D2D formalizes the adversarial process as an evolutionary game in which agents adaptively evolve their rhetorical strategies (Logos, Pathos, Ethos) through competition. To manage the intractably large LLM output space, we decouple agent behavior into a strategic layer, which selects a rhetorical action from a compact strategy space via evolutionary-inspired learning dynamics, and a generation layer, where the LLM produces a concrete passage conditioned on that strategy and the debate context. On four fact-checking benchmarks (FEVER, HealthVer, CLIMATE-FEVER, and SciFact), D2D substantially outperforms existing attacks and evades most defenses.

cs.AI

新论文 · v1arXiv:2611.08757

Goals as a Representational Layer for Human-AI Collaboration

Gaia Molinaro, Lev Tankelevitch, Sean Rintel, Scott Counts

原文摘要

Goals are central to how people, teams, and organizations structure their work. Yet, while generative artificial intelligence (AI) systems increasingly operate across a worker's context, information about why a person is pursuing a task, beyond local prompts and interactions, is typically missing. We argue that goals should become an explicit representational layer mediating interactions among people, AI systems, and organizations. Because human goals are hierarchical, partially latent, and dynamically evolving, such a layer must be multidimensional, both grounded in human articulation and inferred from available evidence, and continuously updated. A goal layer could connect immediate requests to broader outcomes, improve the organization and retrieval of relevant context, support reflection and coordination, and provide a foundation for training, aligning, and evaluating AI systems in terms of human-centered outcomes. We highlight these opportunities while identifying open challenges involving goal inference, socio-cognitive frictions, alignment, privacy, and governance.

cs.HC · cs.AI