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

+197新增论文
+151其中 AI 相关
+16已有论文修订
+91新增 AI 导读

arXiv 最近完成扫描:2026-11-22 08:44 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-21 11:43 — 2026-11-22 11:43(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:2026-11-22 08:44最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 02:59

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

29,158 已收录论文24,181 AI 相关8,992 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 23,345 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-22 08:42(北京时间)
新论文 · v1arXiv:2611.26059

Complete-SVD: A Functional Norm Approach to Low-Rank LLM Compression

Alper Kalle, Theo Rudkiewicz, Mohamed-Oumar Ouerfelli, Mohamed Tamaazousti

原文摘要

Post-training low-rank compression of neural networks must not only approximate the original weights but also preserve the model's input-output behavior. Recent output-aware methods measure the compression error beyond the compressed layer, at the model output, weighting logit changes by the observed next token or by the model's predicted probabilities. We propose Complete-SVD, a method for large language models that minimizes the $L_2$ distance between the logits of the original and compressed models, weighting all logit changes equally. Linearizing the network around the original weights, together with a Kronecker-factored approximation, reduces this objective to a weighted SVD with a closed-form solution. Its output-side factor is intractable to compute; we estimate it without bias from random probes, at a cost of $K$ backward passes per sequence, independent of vocabulary size and sequence length. At matched settings with uniform ranks and WikiText-2 calibration, Complete-SVD achieves the lowest C4 and PTB perplexity in every setting, and at ratio 0.7 under LEMS rank allocation. Replacing the $L_2$ distance by the Kullback-Leibler divergence (KL-SVD), with the same estimation framework, lowers calibration perplexity in most settings and raises downstream accuracy, but raises cross-corpus perplexity. Complete-SVD also improves IFEval instruction-following accuracy over the strongest prior output-aware methods throughout.

cs.LG · cs.CL

版本更新 · v2arXiv:2607.17714

Formal Verification of Continuous-Variable Quantum Programs

Stefanie Muroya, Thomas A. Henzinger

原文摘要

Continuous-variable quantum computing (CQC) is a promising paradigm already accessible through public cloud services. Yet, unlike discrete-variable quantum computing (DQC), CQC lacks a formal program logic. Extending existing quantum Hoare logics to CQC poses fundamental challenges: continuous-variable quantum programs act on {\em infinite-dimensional} Hilbert spaces; their measurement outcomes are often {\em unbounded}, so expectation values may {\em diverge}. We overcome these challenges to give a formal semantics to a universal programming language for CQC and, to our knowledge, provide the first Hoare logic for CQC. We introduce an assertion language based on polynomials over canonical observables and establish soundness and relative completeness of the resulting logic. We also implement a symbolic weakest-precondition calculator for CQC. Our tool has successfully verified CQC algorithms from textbooks and calculated their approximation errors for physically realizable implementations, proved the correctness (i.e., equivalence) of gate decompositions for CQC hardware, and computed the resource requirements (i.e., number of photon-number states) for achieving a desired accuracy in the classical simulation of continuous-variable quantum programs.

quant-ph · cs.LO

版本更新 · v2arXiv:2611.14961

CommJEPA: When Predictive Representation Learning Meets Wireless Intelligence

Feiran You, Hongyang Du, Kang Wei, Jun Li, Abbas Jamalipour

原文摘要

Artificial intelligence (AI) is becoming an integral part of wireless systems, yet much of today's wireless AI remains task-specific and reactive, with models typically trained to map current observations to predefined outputs or reconstruct observed data. While effective for individual tasks, such approaches provide limited support for learning what is predictable, reusable, and transferable in dynamic wireless environments. Joint-Embedding Predictive Architecture (JEPA) offers a different perspective by learning predictive representations in latent space that capture informative structure in the observations. In this article, we introduce Communication JEPA (CommJEPA) as a communication-oriented framework for predictive representation learning. The central idea is bidirectional. JEPA can equip wireless systems with predictive representations for capturing latent channel and environmental dynamics, integrating heterogeneous observations, adapting to changing conditions, and supporting future decisions. Conversely, the physical structure, partial observability, multiple time scales, action-dependent evolution, and resource constraints of wireless systems motivate new communication-aware JEPA designs. We develop this mutual perspective, organize the resulting design space, and discuss representative JEPA-based wireless instantiations for prediction and control. We further highlight open challenges toward scalable, interactive, and predictive wireless intelligence.

cs.NI

新论文 · v1arXiv:2611.26057

SlotFlow-GB: A Neural Framework for Trans-Dimensional Inference of Overlapping Galactic Binaries

Niklas Houba, Giovanni Giarda, Davide Piras

原文摘要

The Laser Interferometer Space Antenna (LISA) will observe the millihertz gravitational-wave sky, including millions of long-lived Galactic binaries (GBs), tens of thousands of which are expected to be individually resolvable throughout the mission's lifetime. Extracting these sources requires jointly determining their number and physical parameters from strongly overlapping signals, posing a computationally demanding trans-dimensional inference problem. We introduce SlotFlow-GB, a neural framework that constructs a learned approximation to the trans-dimensional posterior, producing a probability distribution over source multiplicity and parameter densities for individual binaries in a single evaluation, without requiring their number to be specified in advance. We demonstrate this framework using year-long LISA simulations containing up to ten binaries within a 2 $μ$Hz frequency interval, focusing on a high-SNR regime in which source overlap, rather than individual-source detectability, is the primary challenge. The inferred source count differs from the true count by at most one in 94% of cases, while the model identifies individual binaries even when their signals overlap strongly in frequency. A complete evaluation, including parameter sampling, takes just 41 ms per band on one NVIDIA GH200 GPU, or 530 ms on eight laptop CPU threads. We further demonstrate that these distributions can be used to initialize subsequent likelihood-based inference, connecting fast amortized predictions with joint parameter estimation.

astro-ph.IM · cs.LG · gr-qc · physics.data-an · physics.ins-det

新论文 · v1arXiv:2611.26056

Less Than Meets the Eye: Action-Set Identifiability and Principled Abstention for 3D Physical Geometry

Reshad Ul Karim, Abrar Samin, Sammam Mahdi, Syeda Maliha Tabassum, Md Rashad Al Hasan Rony

原文摘要

A corridor, its mirror reflection, and its image on a display can look identical, yet their physical surfaces lie at different depths. We test whether bounded camera motions distinguish these explanations. Intervene3D takes an image and an initial geometry estimate, predicts the view under each hypothesis and candidate motion, and selects a motion by maximin separability. It reports the mechanism and first physical surface only if the available motions separate the hypotheses still in contention. Otherwise it abstains. An adequacy test also rejects observations that no hypothesis explains. On a matched synthetic benchmark with exact landmarks, disabling abstention gives confident answers on 76.4% of unresolvable scenes. With abstention, none receives a confident answer under exact or noisy landmarks. At 100% accuracy on answered scenes, maximin selection answers 61.0% of scenes and summed selection 43.1%. The adequacy test lowers wrong answers on strongly curved mirror scenes from 56.8% to 12.4%. Real display experiments compare two hypotheses using one frozen depth-network estimate. The second image supplies photometric evidence and is never passed through the depth network. One executed motion reduces the pooled median surface error of twelve networks by 31.5%. These networks have larger error inside the annotated panel than outside on 33.4% of illusion-benchmark images. They place the first physical surface behind the glass on 36.9% of Mirror3D frames.

cs.CV

新论文 · v1arXiv:2611.26055

When Do Rankings Beat Signs? Gradient Direction Recovery under Even Nuisance

Zhuangzhuang Pan, Yan Xia, Chee Seng Chan

原文摘要

Can ranking a batch of antipodal query pairs recover a gradient direction with fewer evaluations than using within-pair signs alone? We study fixed-center recovery in $d$ dimensions for a linear signal plus an unknown differentiable even nuisance, counting function evaluations and feedback rounds separately. Shared pair offsets create an information hierarchy: Mirror retains within-pair signs, Reverse couples reversed cross-pair orders to cancel offsets, and Full enforces all ranking inequalities. Relative shell variation is the nuisance range over the query sphere normalized by the linear-signal scale. When it is at most $c\varepsilon/\sqrt{d}$, Reverse localizes every consistent direction to angular accuracy $\varepsilon$ using $\widetilde{O}(d+\sqrt{d/\varepsilon})$ evaluations in one round at fixed confidence, without a variation bound supplied to the decoder. This matches a lower bound up to logarithmic factors, versus the $Θ(d/\varepsilon)$ one-round sign-only minimax cost on the same class. Over well-conditioned positive-definite quadratics with no common relative-curvature bound, the $R$-round minimax cost is instead $Θ(dR(\varepsilon^{-1/R}-1))$, even with cumulative rankings. We extend validity beyond exact evenness under supplied error bounds. Matched-endpoint experiments certify information beyond signs and short cycles. On two ridge tasks under joint evaluation and round caps, short-cycle recovery attains lower terminal mean gaps than the tested batch-capable controls. An additional optimization benefit from fitting the complete ranking remains unestablished.

cs.LG

新论文 · v1arXiv:2611.26053

Guided Cuckoo Filter: Full-Pipeline Insertion Optimization

Min Liu

原文摘要

Cuckoo filters are widely deployed as space-efficient alternatives to Bloom filters, supporting fast lookups, insertions, and deletions. However, their insertion cost grows steeply at high load factors because the standard insertion procedure is entirely unguided: it places each fingerprint into whichever candidate bucket happens to be tried first, and when both candidates are full, it displaces a random victim in a random walk. These blind random choices occur at two distinct stages of the insertion pipeline, the initial placement decision and the eviction victim selection, yet each stage can be independently replaced with a guided, informed decision. We propose the \emph{Guided Cuckoo Filter}, which optimizes the entire insertion pipeline by steering both stages. At the placement stage, the filter compares the occupancy of both candidate buckets and inserts into the less loaded one, applying the power-of-two-choice principle to reduce how often eviction chains are triggered. At the eviction stage, the filter preferentially displaces the most recently inserted element, whose alternate bucket carries latent randomness that shortens the resulting chain. Because these two guided decisions target orthogonal bottlenecks, eviction frequency and eviction chain length, their benefits compose multiplicatively. Experiments on cuckoo filters with $2^{22}$ slots confirm that the fully guided insertion reduces per-insertion kicks by up to $3.1\times$ at 96\% load and cumulative total kicks by 63\% at the practical capacity ceiling of 95\% load, with negligible computational overhead.

cs.DS

新论文 · v1arXiv:2611.26052

Do Network Mechanisms in Science Vary Across Research Topics? Contextual Relational Hyperevent Models for Large-Scale Bibliometric Data

Jürgen Lerner, Mark Wittek, Raphael Heiberger

原文摘要

Collaboration and citation networks in science are shaped by well-documented mechanisms such as preferential attachment, gender homophily, and self-citation. Whether these mechanisms operate uniformly across research areas, however, is rarely tested, because statistical network models that estimate several mechanisms jointly are difficult to apply to large bibliometric data and to compare across subnetworks. We introduce contextual relational hyperevent models (C-RHEMs), which extend relational hyperevent models by allowing network effects to vary across research topics within a single model. C-RHEMs treat each publication as one event linking its authors to its references, accommodate reference lists spanning several topics, and scale to hundreds of thousands of publications on standard hardware. In a second step, meta-analysis and meta-regression quantify heterogeneity in topic-specific effects and relate it to topic characteristics. We illustrate the approach with about 590,000 oncology publications from the PubMed Knowledge Graph, assigned to 32 OpenAlex research topics. We find that preferential attachment, gender homophily, and self-citation vary substantially across topics, with between-topic standard deviations of 30 to 70 percent of the corresponding pooled effects. Topic age, size, and gender composition account for between a third and two-thirds of this variation: older topics exhibit stronger preferential attachment, self-citation, and gender homophily, and topics with a higher share of male authors exhibit stronger self-citation. The results caution against treating network mechanisms estimated on pooled data as uniform, and C-RHEMs offer a scalable way to study how they differ across research areas.

cs.DL · cs.SI · stat.ME

新论文 · v1arXiv:2611.26050

Studying AI Act Mentions: A first look at the impact of the European Union's AI regulation on Hugging Face

Anamaria Mojica-Hanke, Tessi Aigbovbiosa-Igbinovia, Steffen Herbold

原文摘要

The European Union's Artificial Intelligence Act (AI Act) entered into force in August 2024, and its obligations apply in stages. Obligations for providers of general-purpose AI models have applied since August 2025, and transparency rules since August 2026. These obligations affect developers who share and deploy models and applications on public platforms such as Hugging Face (HF), the largest platform for distributing AI artifacts. However, little is known about how practitioners respond to the regulation in practice. To address this gap, we conducted an empirical study of explicit references to the AI Act on HF. We collected over five million records, including repository cards, discussions, and commits from Models, Datasets, and Spaces, and identified over 1,500 records that explicitly mention the Act. Two coders manually labeled these records using a primarily closed coding approach, and we analyzed the resulting categories per source. We find that explicit references to the AI Act remain rare across all repository types. Repository cards mainly contain claims of compliance and responsibility, like ensuring compliance or assigning responsibility to downstream users, and some claim that the Act does not apply because the model is not a general-purpose AI model. Discussions are used to report alleged violations of the Act, while commits mainly update artifacts that support compliance. Overall, compliance statements focus on who is responsible rather than how compliance is achieved. We believe that this calls for direct platform support, e.g., automated checks and documentation templates, as well as clearer guidance on the scope of the AI Act.

cs.AI

新论文 · v1arXiv:2611.26048

AGE-Capture: Autonomous, General, and Efficient UAV Capture for High-Fidelity Scene Reconstruction

Xiaofeng Jin, Ningbo Bu, Matteo Frosi, Yiran Guo, Weilun Ren, Matteo Matteucci

原文摘要

High-fidelity scene reconstruction with an unmanned aerial vehicle (UAV) requires more than identifying valuable viewpoints: the system must autonomously acquire informative observations and determine task completion under different sensing configurations and limited flight resources. We present AGE-Capture, an autonomous, sensor-adaptive, and efficient active-capture framework for real UAVs. Building on renderability, the framework uses a lightweight online state to evaluate the utility of candidate views, assess task progress, and guide flight execution. By decoupling geometric exploration from reconstruction observations and establishing a unified camera-ray interface, the framework allows different range sensors and RGB camera models to share a reconstruction-driven acquisition pipeline. Reconstruction-frontier viewpoints characterize residual acquisition value for region screening and autonomous termination; dynamics-aware hierarchical planning combined with asynchronous event-driven capture reduces unnecessary inter-region motion, time spent waiting for plans, and repeated observations. Across ten indoor and outdoor simulation scenes, AGE-Capture achieves a mean peak signal-to-noise ratio (PSNR) $3.30$--$3.53\,\mathrm{dB}$ higher than that of the online baselines while using $58.2\%$ of the common mission-time budget on average. Its mean peak GPU memory usage is $91.0\%$--$93.3\%$ lower than that of the same baselines. Experiments with different sensor configurations validate the adaptability of the shared acquisition pipeline, while flights in nine real-world scenes further demonstrate onboard closed-loop operation under practical computational and motion constraints.

cs.RO

新论文 · v1arXiv:2611.26046

Agentic Scaling of Embodied Tasks in Simulation for Synthetic Data Collection and Policy Evaluation

Kang Wang, Xiaolei Zhou, Xiaoyu Wu, Zhendong Wang

原文摘要

Scaling embodied interaction data requires a broad collection of usable tasks, yet specifying and validating each task instance remains costly. We present AgenticTask, an agent-based framework that expands existing task collections under human intent. Its central representation links interaction requirements to object and part bindings, scene and robot configurations, initial conditions, and completion criteria. A task-generation agent reuses interaction structure to construct supported variations and revises incompatible configurations using validation feedback. A task-execution agent uses the same bindings and geometric records to select robot-supported skills and check completion against the original task criteria. On RoboCasa, the framework builds the AgenticTask collection, saving 16,450 task instances from 20,323 construction attempts at a throughput of 166.97 saved instances per elapsed hour, with 62.26% of construction attempts passing the configured physical checks. The execution agent completes approximately 38% of evaluated tasks, with successful cases spanning object placement, articulated manipulation, and sequential arrangement. Diagnostic analysis identifies improvement priorities in construction reliability, action selection, task-robot compatibility, and runtime integration. Together, these results demonstrate task-collection expansion with measured construction productivity and concrete execution evidence, providing a foundation for subsequent interaction-data collection for embodied AI.

cs.RO

新论文 · v1arXiv:2611.26045

PaveSplat: Resolution-Robust Feed-Forward Gaussian Splatting with Residually Corrected Surface-Aligned Covariances

Janghun Kim, Hyun-Cheol Kim, Eunbin An, Myeonghun Oh, Hyon-Gon Choo

原文摘要

We present PaveSplat, a feed-forward Gaussian splatting method that adds a learned residual to the analytic surface-aligned covariance, keeping surfels paving the visible surface across rendering resolutions. Pixel-aligned feed-forward Gaussian splatting predicts one primitive per pixel from a few images in a single pass, and its rendering quality depends on how well these primitives cover the surface. Recent surfel-based methods compute the surface-aligned covariance analytically from predicted depth. However, this covariance has no way to correct the surfel orientation where the depth-derived frame is ill-posed. The surfels then cover the surface loosely; at the training resolution each surfel's one-pixel minimum footprint covers the gaps, but they show up as holes when the same prediction is re-rendered at a higher resolution. To address this, we keep the analytic covariance as an initialization, with its axes aligned to the surface tangents, and add an identity-initialized unit-quaternion residual that reorients a surfel only where this improves rendering, turning surfels that the analytic frame mis-aims back onto the surface. On RealEstate10K, PaveSplat achieves the best PSNR, SSIM, and LPIPS among single-Gaussian-per-pixel methods at every scale (1x-8x). The native-resolution gain is small; the benefit appears when the same prediction is rendered at higher display resolutions without re-running the network. On zero-shot DL3DV, it leads PSNR within this group at every scale and ranks first or second on SSIM/LPIPS. A coverage-hole analysis shows fewer uncovered pixels than both baselines under high-resolution re-rendering.

cs.CV

新论文 · v1arXiv:2611.26044

Variational Trajectory Flow Matching with a Non-Linear Interpolant

Gurjeet Sangra Singh, Frantzeska Lavda, Alexandros Kalousis

原文摘要

Simulation-free methods such as Trajectory Flow Matching learn dynamics by regressing the time derivative of a local interpolant placed between consecutive observations. The interpolant defines the regression target, and under sparse, irregular sampling, it underrepresents the dynamics and bounds what can be learned. We show that no fixed interpolant dominates: the best of linear, quadratic, spline and quintic changes with the system and with the sampling density. We propose to learn the interpolant instead. An amortized implicit neural representation (INR) is fitted to the training trajectories and frozen as a teacher supplying the target velocity to a learnable velocity field. We further frame the velocity field within a variational framework, yielding a probabilistic forecaster that, at inference, requires only the velocity field and no SDE solver. We evaluate our method on a range of dynamical systems: it consistently achieves high accuracy and more stable long-horizon rollouts under sparser observations, at matched velocity-field capacity.

cs.LO

版本更新 · v3arXiv:2408.14443

Ensemble Logic

Guo-Qiang Zhang

原文摘要

Ensemble Logic (EL) is a family of modal logics whose modal parameters are terms with quantified variables. We interpret formulas over metric monoids, that is, commutative monoids equipped with translation-invariant metrics. Displacement evaluates a formula at a translated point, while bounded existence and bounded universality evaluate it at some or every point of a window of prescribed length. Quantification over points and lengths allows the same variable to occur in the parameters of several operators. This provides a common framework for temporal and spatial reasoning in discrete, dense and multi-dimensional settings. Our main result establishes a sharp undecidability boundary for satisfiability over the natural numbers: satisfiability is undecidable already for formulas $\exists N\,\exists n\,Ψ$ in which $Ψ$ is built from the letters of a fixed alphabet by conjunction, disjunction, displacement and bounded universality. We prove this by a reduction from Post's Correspondence Problem. We also present results, in part from companion work, that place this theorem in the general theory of EL. These include an atomic-displacement normal form for covariant windows and composition laws for discrete and dense time. The discrete logic admits an embedding of linear temporal logic, defines the accepting runs of every Büchi automaton, and has a sound equational calculus. For the rational logic, we show that every fixed formula can be evaluated in polynomial time on interval models in which atomic propositions are interpreted by finite unions of intervals. Finally, we represent timed words over positions and real time and compare two window geometries; with rectangular windows, two shared lengths yield undecidability on finite timed words.

cs.LO · cs.AI · cs.FL

版本更新 · v2arXiv:2611.21438

Power Line Detection in Point Clouds with Point-SAM and RANSAC

Christoph Dalitz, Oskar Waedt, Steffen Goebbels, Mahdieh Shamsadini Motlagh, Gabriel Gerzabek

原文摘要

This article deals with the problem of detecting the position of power lines in airborne laser scanning point clouds. We address both the problem of segmenting power lines from background like ground and vegetation and of identifying individual lines. For segmentation we compare a rule-based approach with one based on transfer learning (Point-SAM). For individual line detection, we derive the mathematical formulas for RANSAC applied to 3D catenaries and compare it with a non-parametric method (TriplClust). All methods are evaluated on a newly created dataset "PowLiDFR", which we make publicly available. The samples have been taken from the French project LiDAR HD and cover a variety of terrain layouts and point cloud resolutions. On this dataset, an unsupervised transfer learning outperformed a mere rule-based segmentation, and for individual power line detection, RANSAC was better in following the shape of catenaries than the non-parametric algorithm.

cs.CV

新论文 · v1arXiv:2611.26040

Streaming-AAA: Streaming AAA Game Video Generation at High Resolution, High Frame Rate, and Low Latency

Zhihao Hu, Jianqi Ma, Jun Zhang, Jiayang Bai, Haifeng Shi, Yanxiang Zhuang, Zhongqian Sun, Weidong Zhang 等 9 位作者

原文摘要

Controllable streaming video generation offers a promising route toward interactive game simulation, yet existing methods remain far from practical gameplay. They often fail to jointly achieve high visual fidelity, high frame rates, and low interaction latency, resulting in poor responsiveness and degraded user experience. We present Streaming-AAA, a new paradigm for controllable infinite generation of AAA-quality game videos. Our system generates 1080p gameplay videos at 43 FPS with 198 ms end-to-end latency interaction on merely NVIDIA A800 GPUs, while supporting long-horizon play without severe temporal degradation or collapse. Streaming-AAA is built upon three key designs. First, we redesign the VAE architecture for efficient high-resolution decoding, enabling real-time synthesis at 1080p. Second, we pretrain the diffusion model using diffusion forcing and significantly extend the temporal training horizon, allowing the model to capture long-term dependencies and maintain persistent memory. Third, we perform post-training distillation and memory compression to improve inference efficiency while retaining visual quality and temporal consistency. Experiments show that Streaming-AAA substantially improves visual fidelity, frame rate, latency, and long-term stability over prior controllable video generation approaches, enabling a significantly more playable experience for interactive AAA game video generation.

cs.CV

版本更新 · v2arXiv:2611.23386

Chorus: Shared N-gram Embeddings with Layer-wise Adaptation

Stephen Chung, Wenyu Du

原文摘要

Language models can gain capacity at little computational cost from n-gram embedding tables: large tables of learned vectors for short sequences of consecutive tokens, which are looked up and added to the model's input or to its intermediate layers. Recent designs commonly use a separate n-gram embedding table for each layer, which makes the tables costly in memory and prevents layers from sharing what they learn. However, a single shared table gives every layer the same embedding, leaves no room for layer-wise preferences, and performs poorly in practice. We propose Chorus, in which one set of n-gram embedding tables serves the whole network and every layer reads it in its own way. Each layer maps the refined embedding through its own projection layer, after rescaling its components by factors computed from its current hidden state. Chorus matches the quality of existing n-gram methods with 87% smaller tables: with 1.2B table parameters it predicts text better than FactorEngram and DeepSeek's Engram do with 9.6B. It outperforms Engram, TN-gram and FactorEngram in every setting where we compared them, across table sizes from 20M to 9.6B parameters. We also show that the recently proposed memory-efficient optimizer Ember works well with Chorus, saving 99.2% of table-optimizer memory compared with AdamW without degrading training performance. These results enable further scaling of n-gram embedding tables along the table-size axis by reducing memory use and allowing shared knowledge to be reused across layers.

cs.CL

版本更新 · v3arXiv:2609.20050

The Missing Complement: State-Conditioned Minimal Sufficient Evidence for Coding Agents

Zhexi Feng, Ruiyi Zhang, Yongbo Yang, Pengtao Xie

原文摘要

Retrieval assembles repository context by ranking passages for relevance to the current query. A coding agent halfway through an issue has already read much of what such a ranker returns. Relevance is scored per passage, but sufficiency belongs to the set: independently scored passages can fill the budget with support for one requirement while another goes unmet. We formulate state-conditioned minimal sufficient evidence recovery: given a captured agent state, recover a compact evidence combination supplying what its next decision still lacks. SERBench measures this on 500 held-out states from 45 repositories, recording what the agent has seen, crediting only sets that satisfy every annotated evidence requirement of the current decision, and separating set recovery from candidate discovery. MSS-Complement treats acquisition as set construction, not ranking. Three semantic calls propose a jointly sufficient set, search for what it lacks, and return 4-8 intact source units within 6,144 tokens. One configuration, fixed on calibration data, recovers a complete set for 73.0% of those states at five items and 80.6% at eight, against 61.4% and 72.4% for Qwen3 embedding with reranking. A matched control ranking by similarity alone recovers fewer complete sets, placing the margin over it in the set-level policy, not the computation. The lead persists from frozen repository source with no gold-derived pool. On AMA-Bench it answers from a 76.2% smaller answer prompt, with accuracy 2.08 points above that benchmark's own memory agent. Removing one required group from a complete set costs repair-localization precision under two executors. Retrieval for agents is better posed as recovering what a decision lacks than re-ranking what an issue resembles.

cs.IR · cs.AI · cs.CL

版本更新 · v3arXiv:2603.17927

RoboForge: Cross-space Motion Refinement for Physically Executable Humanoid Whole-body Tracking

Xichen Yuan, Zhe Li, Zhenzhe Zhang, Yangyang Wei, Xinying Guo, Shanghang Zhang, Jianfei Yang

原文摘要

Whole-body motion tracking has rapidly expanded the skill repertoire of humanoid robots, yet nearly all of this progress has come from stronger tracking policies, while the motion references they imitate are treated as fixed supervision. We argue that reference quality is a first-order factor in tracking performance. Human motion data carry capture and estimation artifacts, and retargeting adds physical inconsistencies due to the human-robot embodiment gap; discarding such motions sacrifices valuable skill content, whereas retaining them uncorrected undermines policy learning. We present RoboForge, a data-centric framework that resolves this dilemma by refining references instead of filtering them, repairing each error in the space where it arises. In human motion space, a quality-guided diffusion model localizes corrupted segments and inpaints only those segments, leaving the rest of the motion intact. In robot motion space, kinematic repair corrects foot-contact and self-collision defects, and annealed sampling-based optimization under physics rollouts turns kinematically plausible references into dynamically executable ones. RoboForge requires no change to the tracker: it improves three retargeting methods and four tracking policies, raising the success rate of BeyondMimic from 89.51% to 96.20% while reducing the mean per-joint position error (MPJPE-L) by 31.6%. Refined references also accelerate policy learning, restore tracking performance from heavily corrupted inputs, and enable more than 30 high-dynamic motions, including flips and the Thomas flare, on a real Unitree G1.

cs.RO

新论文 · v1arXiv:2611.26035

Fixing the Twin, Not the Detector: Sim-to-Real Roadside LiDAR Detection from Digital-Twin Data

KuanLin Chen, Feng-jui Kuo

原文摘要

Roadside LiDAR detectors are expensive to train on real labels, and digital twins promise to replace these labels, but a detector trained on a twin meets a different sensor, different objects and different annotation conventions on the real site. We describe the pipeline that placed first in the LUMPI track of the UrbanTwin Sim2Real LiDAR Challenge (IEEE BigData 2026 Cup, 0.5541 combined score) and fifth on its V2X-Real track (0.8102). Every detector is a CenterPoint model trained only on synthetic data, on a single 8 GB laptop GPU. Most of the gain came from correcting differences between the twin and the real site rather than from the detector. We align the twin's coordinates, point density and box convention with the test sensor, generate the pedestrians and riders that the twins lack, plus street clutter as unlabeled negatives, and rebuild twin objects at real size and real point density. A non-learned resampling to the per-cell point density of unlabeled real frames lifts LUMPI realism from 0.589 to 0.947, against 0.949 between real frames. Label-free post-processing then exploits what a fixed roadside sensor guarantees: static structures recur in every frame, and real objects have characteristic point statistics. Reporting the validation and test gain of each component side by side, we find that corrections of systematic sim-to-real differences transfer to the test set, typically keeping a quarter to three quarters of their validation gain, while choices that fit the few objects of a validation set do not.

cs.CV · cs.LG

新论文 · v1arXiv:2611.26033

SEOYUN: Refusal-Inducing Exam Documents for AI-Glasses Pipelines

Dohyeon Kwon

原文摘要

Camera-equipped glasses can convert an examination page into a query to a vision-language model (VLM). Device detection addresses possession; a complementary question is whether an examination document can reduce the usefulness of an unauthorized capture. We present SEOYUN, an application-specific framework for refusal-inducing exam documents, with optical and encoder-level interventions as extensions. We use reverse jailbreaking to describe the direction of the objective: inducing abstention rather than bypassing refusal. Refusal induction and defensive visual prompt injection already have precedents; our contribution is a threat model, analytical limits, and an evaluation protocol for examination use. We distinguish refusal from correct-answer leakage, show why removing an added dummy option is not itself a reduction in the original answer space, and derive a sampling comparison under explicitly stated camera assumptions. The protocol separates digital, simulated, and physical inputs and requires held-out evaluation after subject-specific selection. A local CSAT candidate corpus and rendering harness have been prepared, but this version reports no model-performance measurements, physical validation, or human-subject results. Accordingly, SEOYUN is a research framework, not a validated defense or a basis for allegations against individuals.

cs.AI

新论文 · v1arXiv:2611.26031

Close&Safe: Maximal Task Preservation for Redundant Manipulators via C-Space Boundary Detour

Abhigyan Singh, Manthan Goyal, Fares Abu-Dakka, Ravi Prakash

原文摘要

Contact-rich manipulation such as wiping, polishing and surface inspection specifies the task as a continuous end-effector trajectory rather than a goal pose. When an obstacle absent from the planning model intersects it, conventional planners return a collision-free path that may differ from the nominal throughout, discarding trajectory that was never obstructed. We present Close&Safe (Close to the intended task, Safe from obstacles), which decouples tracking from avoidance: the nominal is followed exactly wherever it remains feasible, and avoidance acts only on the interval where it is not, so tracking is hindered as little as the obstacle allows. On that interval we sample the configuration-space obstacle boundary within a corridor around the nominal, connect the samples into a roadmap, and search it for the detour that deviates least before rejoining the nominal at the earliest feasible configuration. We evaluate across diverse simulated obstacle configurations, physical Franka experiments, and ablations of key algorithmic components. The results demonstrate collision-free trajectory repair with localized task modification, while recovering 52.4-78% in simulation and on hardware.

cs.RO

新论文 · v1arXiv:2611.26030

The Delayed Societal Cost of Unfairness

Giovani Valdrighi, Isabel Valera, Marcos Medeiros Raimundo

原文摘要

Fair machine learning typically evaluates the societal costs of a decision policy -- the loss incurred by the decision-maker (DM) and the unfair distribution of costs among demographic groups of decision-subjects (DS) -- on static data, where reducing one often comes at the expense of the other. Yet decisions change the populations on which they act, and prior work has shown that fairness interventions designed on static data may fail to benefit, or even harm, the disadvantaged group over time. We investigate a complementary question: When can a policy that is Pareto-optimal on pre-deployment data be outperformed by a fairer policy once each is evaluated on the population it induces, reducing costs for one stakeholder without increasing them for the other? In a setting where qualification scores evolve in response to binary decisions and their outcomes, we show that under Rawlsian fairness, which never lowers any group's acceptance rate, placing sufficiently high weight on fairness reduces societal costs whenever accepting disadvantaged DS at the DM's break-even point leaves them in a state less costly to the DM than rejecting them. Under egalitarian fairness, which may lower the advantaged group's acceptance rate, this additionally requires the reduction in DM loss on the disadvantaged group to outweigh the increase on the advantaged group. Under both fairness notions, however, when DM loss is accuracy-based, improvements in the post-deployment population may arise from declining DS qualification, revealing that gains for the DM do not necessarily translate into better outcomes for DS. Our results show that policy-induced population dynamics can overturn apparent performance-fairness trade-offs, allowing fairer policies to reduce the delayed societal costs of decision-making.

cs.LG

新论文 · v1arXiv:2611.26029

Serialization Tax in Shared-Latent Exchangeable Decisions

Siming Zhang, Zhehui Shen, Shijie Chen, Xinle Gu, Yansen Yu, Hang Yu

原文摘要

Accurate item probabilities can produce inaccurate group-level decisions. A system may score each item separately, yet act on whether enough hidden items succeed: show a slate, defer a panel decision, or inspect a batch. When items share an uncertain latent state, independent marginalization preserves means but removes dependence from the aggregate posterior. We call this error the serialization tax. For binary items sharing one success rate, classical convex-order comparisons give a more dispersed exact count law than any independent completion with the same total mean. We quantify variance and total-variation gaps, derive threshold separation under margin conditions, and connect posterior approximation to decision regret. Separate results cover non-binary outcomes, heterogeneous items, and sufficient posterior resolution. TaxScore identifies large hidden blocks with unresolved uncertainty near a decision boundary for routing to richer posterior heads. On pre-specified, hidden-target-free MovieLens 1M/10M audits, existing shared-latent heads improve high-score negative log-likelihood (NLL)/utility from 0.673/0.696 to 0.629/0.723 and from 0.670/0.691 to 0.639/0.715. Routing gives 0.006--0.012 full-population utility gains, with matched controls identifying their sources. Our code is available at https://github.com/SiminChang/Serialization-Tax-in-Shared-Latent-Exchangeable-Decisions.

cs.AI

新论文 · v1arXiv:2611.26024

ForeSeek: Foresight-Driven Reasoning for Long-Form Video Understanding

Yuchen Sun, Siyuan Cheng, Shi-Zhe Chen, Yixin Qin, Jinwen Luo, Zheng Wei

原文摘要

Long-form video understanding, as a fundamental yet crucial vision task, is being revolutionized by multimodal large language model (MLLM)-based video agents. However, the retrieval-driven nature inherent to these agents creates a significant bottleneck, forcing them to reactively fetch clips based on local saliency without a global temporal plan. To address this, we present ForeSeek, a foresight-driven video agent that transforms long-video reasoning from reactive retrieval into predictive exploration. Rather than blindly probing the timeline, ForeSeek forms visually grounded expectations over unseen temporal regions, verifies them through targeted observations, and continuously refines its search based on emerging evidence, where failed expectations serve as informative signals for principled backtracking. Furthermore, we develop a two-stage training framework that equips MLLMs with foresight-guided reasoning and efficient decision making. Extensive experiments on five long-video benchmarks demonstrate that ForeSeek consistently improves the efficiency--accuracy trade-off of long-video reasoning, achieving competitive or superior accuracy while observing $2.5$--$4.4\times$ fewer frames than the strongest search-based agent. Code is available at https://github.com/harry0021/ForeSeek

cs.AI

版本更新 · v2arXiv:2611.17309

Classical Energies Learn to Reason 1,000x Faster than Looped Models

Benjamin Hoover, Hugh Blayney, James C. R. Whittington

原文摘要

Looped models have emerged as a prominent paradigm for reasoning, with successes on simple reasoning puzzles such as Sudoku-Extreme and Maze-Unique. However, these models can be unstable, slow to train, and use many parameters and FLOPs to solve simple puzzles. Descent on an energy not only mitigates these issues but also offers advantages: built-in early stopping, compute that adapts to problem difficulty, fast training without differentiating through a solver, and interpretable weights that learn the constraints of the puzzle. Here we demonstrate this with a learned version of Hopfield & Tank's energy from the 80s, which we term CRAM, solves 99.996% of Sudoku-Extreme and 100% of Maze-Unique test puzzles after 42 ms and 132 ms of training on one GPU, at least 1,000x faster than looped models on these tasks. CRAM also matches baseline looped models at their highest accuracy on Sudoku-Extreme using at least 282x fewer inference FLOPs. Our work suggests two things. First, that energies are an extremely promising direction for training reasoning from data. Second, a caution against using trivially solvable benchmarks like Sudoku and Maze to test the scalability of looped reasoning techniques.

cs.AI · cs.LG

版本更新 · v2arXiv:2610.35376

Converse symplectic interlacing via low-rank singular value perturbations

Priyanka Grover, Hemant K. Mishra, Sarishti Singh, Temjensangba

原文摘要

The symplectic interlacing theorem states that the symplectic spectrum of a $2n \times 2n$ real symmetric positive definite matrix satisfies a $2k$-step interlacing relation with the symplectic spectrum of every $2(n-k) \times 2(n-k)$ $s$-principal submatrix. We prove the converse in this paper. More precisely, given a global symplectic spectrum of size $n$ and a local symplectic spectrum of size $n-k$ satisfying the $2k$-step interlacing inequalities, we construct a $2n \times 2n$ real symmetric positive definite matrix having the prescribed global symplectic spectrum and a $2(n-k) \times 2(n-k)$ $s$-principal submatrix of it having the prescribed local symplectic spectrum. A novelty of our method is building a connection between the $s$-principal submatrix problem and low-rank singular value perturbations. We also provide an algorithm to determine a positive definite matrix that realizes the given global and local symplectic spectra. In the context of the Gaussian quantum marginal problem, our result implies that the symplectic interlacing inequalities are not merely necessary conditions on Gaussian marginals, but are also the complete spectral compatibility conditions for the global-versus-local-mode marginal problem.

math.FA · cs.IT · math-ph · math.SG · math.SP

新论文 · v1arXiv:2611.26020

FrontierReplay: Split-Free Exact Prefix Caching for Hybrid Language Models

Zekun Wang, Yizhuo Wang

原文摘要

Hybrid language models combine self-attention with recurrent layers whose in-place state updates restrict prefix reuse to exact saved positions. Saving states on a coarse grid wastes reusable prefixes, whereas existing fine-grained exact caches must cut the prefill forward pass at each mid-prompt checkpoint, adding scheduler steps that outweigh the reuse gains under load. We observe that across multi-turn chat, agent loops, document question answering, and tree search, prefix reuse almost always ends at structural boundaries predictable in advance, such as turn ends and known branch points. We present FrontierReplay, a prefix caching system that materializes exact checkpoints at these boundaries without cutting the prefill. While the forward pass runs uncut, FrontierReplay logs a short window of recurrent-layer inputs and invokes a cross-layer, state-only replay kernel to build the target state 6-37x cheaper per token than full recomputation without host synchronization. A frontier-aware retention policy prioritizes active anchors and proven branches while lending dead checkpoint slots to the attention KV cache. Across diverse hybrid architectures, model scales, and workloads, FrontierReplay delivers clear gains over existing exact caches, lowering median time to first token (TTFT) by 16-89% (up to 8.9x), raising 80% service-level objective (SLO) capacity by up to 1.73x, and improving goodput by up to 4.26x (3.67x at larger scales). Compared with approximate replay methods, FrontierReplay delivers up to 1.54x higher goodput while preserving exact task accuracy, avoiding the up to 17% accuracy drop suffered by approximation on long contexts.

cs.LG

新论文 · v1arXiv:2611.26019

A Task-Level Bias Taxonomy for Historical Image Retrieval

Omkar Gavali, Ratan Sebastian, Christoph Rippe, Ralph Ewerth, Anett Hoppe

原文摘要

In recent decades, galleries, libraries, archives, and museums (GLAM institutions) have digitised vast amounts of cultural heritage material, including historical imagery. A major challenge in making these collections publicly accessible is enabling effective search and discovery. Many historical images contain incomplete, sparse, or no descriptive metadata. Consequently, machine learning (ML) methods are increasingly being used to generate or enrich metadata to improve the searchability of such collections. However, artificial intelligence (AI)-based systems can reproduce or amplify existing biases, particularly when applied to historical imagery. Factors such as differences between the data used to train the models and the historical images on which they are applied, degraded image quality, monochrome photographs, and historically biased or incomplete descriptions can reduce model performance and introduce bias into search and retrieval results. In this paper, we examine different forms of bias in historical image collections and review current methods to mitigate them. We investigate how these biases arise and how they influence AI-supported search and retrieval systems.

cs.IR

新论文 · v1arXiv:2611.26012

When Agents Play Differently: Localizing Human--Agent Behavioral Divergence with Conditional Distributional Testing

Oleg Smirnov, John Pertoft, Sara Karimi

原文摘要

Evaluating whether artificial agents behave like humans requires comparing not just average outcomes but entire behavioral distributions across diverse contexts. We propose a stratified framework that computes nonparametric discrepancy statistics within contextual cells, and discovers subgroup rules identifying where agent behavior departs from human play. We show that session-level metrics dominated by win/loss outcomes offer little signal beyond mean comparisons, whereas process-level metrics such as move efficiency reveal genuine behavioral differences. Applied to a commercial match-3 game, it uncovers patterns invisible to aggregate metrics: a playtesting bot is diverging substantially from humans in certain game modes, while showing no detectable divergence in others. We release our implementation as open source to support reproducibility and adoption.

cs.AI