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

+479新增论文
+361其中 AI 相关
+64已有论文修订
+93新增 AI 导读

arXiv 最近完成扫描:2026-11-16 01:34 · 全站最近入库:2026-11-16 21:56各信源进度 →

统计口径与数据新鲜度

统计 2026-11-15 22:57 — 2026-11-16 22:57(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-16 01:34最近发现新增版本:2026-11-16 21:56库内论文最新更新时间:2026-11-14 02:59

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

23,827 已收录论文19,941 AI 相关6,922 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 23,827 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-16 01:33(北京时间)
新论文 · v1arXiv:2611.11547

Dropping Tokens Without Dropping Them: Causal Attribution for Diabetic Retinopathy

Pramath Girish Haritz, Chandra Sekhar Seelamantula

原文摘要

In this work, we investigate whether attention-based attribution maps in a vision transformer provide causal explanations for diabetic retinopathy (DR) classification. We introduce a principled decomposition of attention into two attribution maps: the contribution ($c_t$), which satisfies the completeness axiom in explainable artificial intelligence (xAI), and the pull ($p_t$), which quantifies causal explanations. We define an identity relating $c_t$ and $p_t$ that characterizes when $c_t$ fails to capture causality. We test whether each map is causally faithful on the MAPLES-DR dataset by dropping/inserting tokens, and report Necessity ($N$), Sufficiency ($Su$), and Faithfulness Correlation ($FC$). The pull $p_t$ achieves the highest causal alignment ($FC = 0.53$, $N = 0.71\%$, $Su = 3.0\%$), outperforming Attention Rollout, Grad-CAM, and AttnLRP. Occlusion retains better sufficiency ($Su = 1.6\%$) but is about 800 times slower. Furthermore, on MAPLES-DR, $p_t$ achieves $70\%$ of the average precision (AP) of HRDecoder, a state-of-the-art supervised lesion segmentation model, despite the absence of pixel-level supervision. Together, these findings show that the pull $p_t$ and the contribution $c_t$ form computationally efficient attribution maps for reliable causal interpretation in safety-critical medical diagnostics. The code and demo are available at https://github.com/pramath-haritz/dr-xai.

cs.CV

版本更新 · v2arXiv:2603.05625

SHERLOC: Statistical Hypothesis Evaluation and Recovery for Learning Opponent Characteristics

Soyon Choi, Scott Alfeld, Meiyi Ma

原文摘要

When used in automated decision-making systems, machine learning (ML) models are vulnerable to data-manipulation attacks. Some defense mechanisms (e.g., adversarial regularization) directly affect the ML models while others (e.g., anomaly detection) act within the broader system. However, many defenses operate under a predefined threat model that assumes characteristics of the adversary, including their knowledge, capabilities, and objectives, are known and fixed. In practice, these characteristics may be unknown or evolve over time, creating a mismatch between the assumptions of the defender and the true attacker. In this paper, we study a new perspective on adversarial defense: focusing on the attacker, rather than the attack. We present and demonstrate SHERLOC, a framework for adversarial learning forensics -- namely, a detective that learns about the attacker from observed attack(s). We prove that, without additional knowledge, the attacker is non-identifiable --- multiple potential attackers would perform the same observed attack. This reveals an important distinction between reconstructing an attacker's behavior and recovering the attacker that generated it. We demonstrate empirically that SHERLOC is able to reconstruct a hypothesis attacker that is consistent with the observed attacks against various defender models (i.e., generalized least-squares linear regressors, logistic regressors, and polynomial classifiers) across various security-critical domains. Finally, we show that these inferred attackers need not match the ground truth attacker exactly to have downstream utility. Defenses using the SHERLOC-inferred attackers achieve approximately 90-100% of robustness improvement that would be achieved if the defender had access to the ground-truth attacker.

cs.LG

新论文 · v1arXiv:2611.11542

Dashboard Remixing: Goal-oriented Organization and Enrichment of Visualizations and Insights

Junran Yang, Leilani Battle, Jane Hoffswell

原文摘要

Dashboard creation requires iteratively authoring multiple views and organizing them into a structured format. However, creators often struggle to step back and assess their own work for organization and refinement due to cognitive limitations. In this work, we propose a pipeline to provide computational support for self-reflection during the design process. Our pipeline automatically infers analytical objectives, discovers implicit patterns, and structures those patterns to surface coherent analytical pathways. It further enriches existing dashboard designs by generating complementary visualizations and aggregating natural language insights based on relationships across data entities (e.g., contrasting, drill-down, and relatedness) to reveal chart connections. Our rule-based pipeline provides objective, verifiable feedback, offering an external perspective while enabling creators to tell their own data stories. Four case studies and six expert interviews demonstrate how this approach helps creators reflect on organizational alternatives and surface hidden analytical opportunities in a principled way, reducing the cognitive burden while encouraging deeper analytical exploration.

cs.HC

新论文 · v1arXiv:2611.11541

Safeguarding Agentic AI in Radiology Workflows: A Multimodel Evaluation of Rule-based and Large Language Model-based Verification

Semil Eminovic, Robin Schmidt, Bogdan Levita, Maximilian Lindholz, Tobias Penzkofer, Jawed Nawabi

原文摘要

Purpose: To estimate how often tool-using large language model (LLM) agents record clinically relevant false or unsupported information in simulated radiology documentation and how much safeguards applied before each action reduce these errors. Materials and Methods: In this simulation study (October 2026), 48 synthetic, radiologist-reviewed base scenarios covered finding communication, tumor board preparation and follow-up management. Forty-three also had a misleading-information variant (91 variants). Six LLM agents (Claude Sonnet 5.5, GPT-6.1-Sol, GLM-5.3, MiniMax M3, Gemma-4-31B, Qwen3.8-27B) completed each variant with permissions only, deterministic rules, and rules plus a locally run LLM checker (1638 episodes). The primary outcome was at least one executed high-severity violation per episode. Risk differences were estimated per agent at base-scenario level and, exploratorily, pooled across agents, with analyses of document access and blocks. Results: With permissions only, 7.3%-27.1% of episodes had a high-severity violation (pooled across agents, 19.4%). Pooled, it was 16.7% with rules (-2.8 percentage points; 95% CI: -6.1, 0.5) and 13.2% with rules plus the LLM checker (-6.2; 95% CI: -9.5, -3.0). Adding the LLM checker to rules changed it by -3.5 (95% CI: -6.3, -0.7). No difference between agents was detected (P = .43), and task success did not decrease detectably. In misleading variants, agents opened the altered document in 83% of episodes regardless of error, and 58% of erroneous statements cited it. Most blocked actions (64%-68%) would not have caused a violation and were returned for revision. Conclusion: Layered safeguards reduced but did not prevent high-severity errors. Errors after misleading information reflected misjudged rather than missed information.

cs.AI · cs.CL · cs.CY

新论文 · v1arXiv:2611.11540

Hydrograd v2.0: a differentiable two-dimensional shallow-water model on unstructured meshes with steady and unsteady adjoints for calibration and closure learning

Xiaofeng Liu

原文摘要

We present \textit{Hydrograd} v2.0, a significantly improved differentiable 2D shallow-water model written in Python with JAX. The model solves the shallow-water equations on the unstructured meshes of production models such as SRH-2D and HEC-RAS 2D with a well-balanced, positivity-preserving finite-volume scheme that handles wetting and drying. Gradients are tailored to the type of problem. Steady states are found by Newton's method with an exact sparse Jacobian obtained by graph colouring, and their gradients follow from the implicit function theorem at the cost of one linear solve. Unsteady flows are differentiated in reverse mode with checkpointing, or with an implicit time integrator whose gradient costs about two forward runs. All gradients agree with finite differences, to about $10^{-8}$ for steady flows. On the Savannah River case, the steady water surface differs from that of SRH-2D by 4.7\,mm (root mean square). Steady closure learning was 96 times faster than differentiating through time marching and reached a lower loss. An inflow hydrograph was inverted from stage records with a peak error of 2\,\%. In an unsteady roughness calibration, a singular value decomposition analysis of the sensitivity matrix showed that the stage records at the gauges are 70 to 160 times more sensitive to the total resistance of the reach than to its distribution among the roughness zones, which explains why only a gauge layout with gauges in every zone recovered all five roughness values. On one GPU, a mesh of 512,000 cells takes 7.6\,s for a forward run and 51\,s for a gradient.

physics.flu-dyn · cs.CE · physics.comp-ph

新论文 · v1arXiv:2611.11539

PRISM: Causal Counterfactual Minimization of Agent Tool Privileges via Information Bottleneck Role Induction

Yang Luo, Qingni Shen, Tairui Wang, Jiajun Li, Zhonghai Wu

原文摘要

This paper shifts the object of role engineering from human users to large language model (LLM) agents, and proposes PRISM to address the over-authorization that arises when an agent obtains full tool privileges through the Model Context Protocol (MCP). PRISM formulates whether a tool is necessary for a task as an intervenable causal proposition, and applies structured interventions in a sandbox to learn a counterfactual necessity model. It then compresses trajectories into discrete latent variables through an information bottleneck, and determines the number of roles from the knee of the rate-distortion curve. Finally, it formalizes pruning as a bilevel Stackelberg game, in which a generative flow network samples diverse near-optimal solutions. On three public benchmarks, AgentDojo, $τ$-bench and ToolEmu, PRISM compresses the number of authorized tools per role by 78.3%, attains a task success rate of 78.2%, reaches a counterfactual necessity accuracy of 88.7%, and lowers the blast radius to 15.9%.

cs.CR · cs.AI · cs.LG

版本更新 · v2arXiv:2611.06321

EnzymeRSI: A dual-loop recursive self-improvement system for enzyme design

Genoria AI Team, Yuejie Hou, Lixiang Yang, Yisi Liu, Weiting Tang, Yueyuxiao Yang, Meng Yang

原文摘要

Enzyme engineering usually optimises which candidates to assay within a fixed model and workflow; when the bottleneck is a mismatched method or a failing workflow, that route is hard to escape. We present EnzymeRSI, a dual-loop system for recursive self-improvement (RSI). The outer loop evolves harness assets without changing model weights; the inner loop develops prediction and selection algorithms under controlled conditions; a cloud laboratory system is specified for closing the physical loop with quality-controlled feedback. The outer loop raised the mean development-set Spearman correlation on ProteinGym by 0.044 across five assays, reaching 0.335 on CD19 and 0.683 on OPSD (rhodopsin); of the two first places, only CD19 was produced by the outer loop itself. In ThermoMPNN self-improvement the inner loop reached a PCC of 0.849, and on TadA-Bench a Spearman correlation of 0.307 in 1.13 h. We examined the system's experimental decision-making in three enzyme systems. For the luciferase RLuc, 10 of the 32 nominated substitutions not pre-specified reached at least twice wild-type activity in a single round. For the cytidine deaminase A3A, a variant selected by a rule that caps loss of unmodified C raised 5mC conversion in the weaker CT context from 11.4% to 46.5% and reversed the substrate preference in both contexts. For the DNA polymerase KOD-E, 189 single-substitution variants defined a landscape on which different objective functions selected divergent next steps from the same parent; one assayed combination reached 31.1 times the parent on a modified substrate. By treating the organisation of research and the development of algorithms as separately scored objects of evolution, and linking a frozen decision record to a physical measurement, this work offers an evolutionary paradigm for moving beyond the blind spots of a fixed method and for autonomous enzyme engineering.

cs.AI

新论文 · v1arXiv:2611.11537

Weight-Space Programming of Variational Quantum Circuits: Large Language Models Write Quantum Circuit Parameters Without an Optimizer

Debarshi Kundu, Swaroop Ghosh, Vasant G Honavar

原文摘要

Variational quantum algorithms (VQAs) optimize circuit angles through repeated quantum-device evaluations, often facing vanishing gradients and high optimization costs. We ask whether a large language model (LLM) can write these angles directly. Extending *weight-space programming* from classical networks to quantum circuits, we provide a frontier LLM (\gpt) with the problem, circuit, and gate conventions. Its angles are executed exactly as written, without returning losses, gradients, or accuracy feedback. Across 139 requests, we test handwritten-digit classification, VQE ground-state estimation for spin chains and H$_2$, and MaxCut QAOA, comparing against random weights, classical optimization, and optimization initialized with LLM weights on the same circuit. When parameters have interpretable structure, the LLM matches or approaches the optimizer: it achieves 95--100\% of the energy spectrum on every VQE problem, derives the exact H$_2$ ground state in closed form, finds optimal depth-one QAOA angles for every graph, and reaches 93--96\% accuracy on digits 3 versus 8 with 12--18 hand-written weights, outperforming an optimizer trained on the same ten examples. Its weights are also strong warm starts: optimization reaches its best value after a median of one evaluation rather than hundreds. Allowed to design circuits, the LLM chooses data-dependent rotation chains encoding pixel importance in all 30 cases (one qubit for binary tasks, one chain per qubit for ten classes), reaching 65\% accuracy across ten digits. Yet the approach fails in opaque weight spaces: accuracy drops to chance when fixed circuits exceed 30 weights, while optimization improves. It is also sensitive to gate conventions: 6 of 42 QAOA replies reverse mixing-angle signs and fail. We release all prompts, replies, and scores.

quant-ph · cs.AI

新论文 · v1arXiv:2611.11535

Addressing Bias Early in the AI Lifecycle: Social Determinants of Health in Functional and Cognitive Assessment Labels of Nursing Home Residents

Isabel Amaya-Rodríguez, Cristina Martin, Nekane Larburu, Teresa Garcia-Navarro, Roberto González Novas, Alberto Rodríguez Taboada, Maite García-Sebastian, Ainara Estanga 等 13 位作者

原文摘要

Fairness in clinical artificial intelligence (AI) is usually evaluated after model training, yet bias can enter earlier in the AI lifecycle, during data collection and labelling. When routine clinical assessments serve as target variables, systematic differences in these labels across sociodemographic groups are carried into model development, validation and updating. This study addresses this early lifecycle stage by auditing labels from three standardised instruments used in residential care: the Barthel Index (BI), the Mini-Mental State Examination (MMSE) and the Global Deterioration Scale (GDS). Using electronic health records of 2,407 residents from 12 nursing homes in the Basque Country, Spain (2011-2021), we assessed four social determinants of health (SDoH) - sex, educational attainment, lifetime work activity and marital status - with two group fairness metrics, Disparate Impact (DI) and Statistical Parity Difference (SPD), computed for each outcome category with bootstrap confidence intervals. Disparities were found for all four attributes and across the three instruments: men, residents with intermediate education and never-partnered residents were more frequently classified in categories indicating preserved function or cognition. These disparities may reflect genuine differences in health status, measurement properties of the instruments, assessment practices or selective admission to residential care, and cannot be attributed to a single cause with the present design. Regardless of their origin, any model trained on these labels would inherit them, and they could persist through later lifecycle stages unless explicitly examined. These findings support pre-training label auditing as the first step of a continuous fairness assessment spanning model development, validation, deployment and updating.

cs.LO · stat.OT

新论文 · v1arXiv:2611.11533

Ostinato: Fast CPU-GPU MoE Inference for Local Agents via Resident Experts in Every Forward Pass

Junning Zhu

原文摘要

Language-model agents increasingly run on workstations with a few consumer GPUs, where large mixture-of-experts (MoE) models keep part of their experts in host memory. With existing CPU-GPU engines, these agents get fast prefill or fast decode, not both: llama.cpp's expert cache skips forward passes of more than 32 tokens to keep the experts of decode cached, so an agent waits 2.1-2.8 s after each tool output, whereas its default layer placement shortens this wait but is 1.5-1.6$\times$ slower in decode. Our key insight is that a short prefill touches too many distinct experts to upload, yet sends 84% of its expert activations (routes) to a stable set of experts that fits in GPU memory. We present Ostinato, a CPU-GPU MoE inference engine that computes these experts on the GPUs in every forward pass without changing what the model computes. Short prefills split their routes, leaving fewer than one of a token's six routed experts per layer to the CPU; long prefills upload each distinct cold expert once; and each decode step runs as one fixed-shape CUDA graph per GPU. On real coding and customer-service agent sessions with DeepSeek-V4-Flash on four RTX 4090D GPUs, Ostinato is the fastest engine in both phases and finishes every session sooner than all six baselines. Against llama.cpp and KTransformers, it cuts the median wait after a tool output by 22-55% and finishes sessions 1.4-2.5$\times$ sooner.

cs.DC · cs.LG

新论文 · v1arXiv:2611.11532

DriveFoundry: Co-Evolving Data and Model for Efficient Closed-Loop Iteration in Autonomous Driving

Ziyang Yao, Yuncheng Jiang, Fang Li, Haochen Liu, Shaoqing Xu, Wen Zhang, Xiangyang Chu, Zibin Guo 等 27 位作者

原文摘要

A driving foundation model must learn reusable decision knowledge from rare critical events and apply it under strict runtime constraints. We introduce DriveFoundry, a closed-loop framework improving iteration efficiency under data and computation budgets by coupling structured causal data refinement with adaptive inference. The Data Engine builds traceable scene-candidate records linking scene evidence and factor-decision relations to candidate trajectories and outcome assessments. Candidate comparisons and counterfactual annotations clarify how scene factors affect behavioral feasibility and preference. Verified model failures guide VLM-assisted annotation and expert correction, refining training data and annotation systems. The model engine, Xiaomi-Drive-o1, unifies Perception, Causation, Action, World, and Value within a shared vision-language backbone to enable cross-task knowledge transfer. Block-wise latent refill passes candidate-specific evidence and predicted visual consequences between task blocks, connecting scene understanding, planning, and evaluation. A learned value-of-computation head selects tasks and allocates latent tokens based on expected benefit and cost, adapting inference to each decision. On PhysicalAI, the 9B model achieves the lowest ADE@1 across evaluated splits and horizons among tested models. In AlpaSim, it outperforms Alpamayo 1.5 across all metrics and achieves lower close-encounter and off-road rates than the 34B Alpamayo 2 Super, while trailing in progress and AlpaSim scores. For the 0.8B variant, joint training reduces 5 s minADE@6 by about 54% without altering the inference path or token budget. Adaptive inference matches full reasoning performance using only 28% of tokens. Real-vehicle tests confirm runtime feasibility, supporting a unified paradigm for learning from critical experience and dynamically allocating computation.

cs.AI

新论文 · v1arXiv:2611.11531

Probing Transformer Representations Layer by Layer, Module by Module

Ambroise Odonnat, Theo Gnassounou, Laetitia Chapel, Romain Tavenard, Ievgen Redko

原文摘要

Which representation should be probed in a transformer-based foundation model? Recent work shows that the standard practice of probing the final layer is often suboptimal, as intermediate layers can provide better features. Yet, representations are still largely studied along depth alone, treating each transformer block as an atomic unit. In this work, we analyze transformer representations layer by layer, module by module. Through more than 144,000 linear probing evaluations of vision foundation models (VFMs) and large language models (LLMs), we show that layer and module choices should be considered jointly: the best layer-module pair yields substantially larger gains over the final-layer residual stream than the best layer with the residual stream fixed. Moreover, module performance can vary substantially across layers, showing that useful representations cannot be characterized by depth alone. Building on these findings, we evaluate simple fixed probing strategies: the final-layer feedforward activation (Act@100) for vision and the second normalization layer at 60% depth (LN2@60) for language. These strategies improve over the probing baseline in a majority of evaluations and recover, on average, 42% and 86% of the gains from probing the best layer-module pair for vision and language, respectively. For VFMs, these gains increase with the transfer difficulty between pretraining and downstream data.

cs.LG · cs.AI

版本更新 · v3arXiv:2506.15018

Private Continual Counting of Unbounded Streams

Ben Jacobsen, Kassem Fawaz

原文摘要

We study the problem of differentially private continual counting in the unbounded setting where the input size $n$ is not known in advance. Current state-of-the-art algorithms based on optimal instantiations of the matrix mechanism cannot be directly applied here because their privacy guarantees only hold when key parameters are tuned to $n$. Using the common `doubling trick' avoids knowledge of $n$ but leads to suboptimal and non-smooth error. We solve this problem by introducing novel matrix factorizations based on logarithmic perturbations of the function $\frac{1}{\sqrt{1-z}}$ studied in prior works, which may be of independent interest. The resulting algorithm has smooth error, and for any $α> 0$ and $t\leq n$, it is able to privately estimate the sum of the first $t$ data points with $O(\log^{2}(t)(\log\log t)^{1+2α})$ variance. It requires $O(t)$ space and amortized $O(\log t)$ time per round, compared to $O(\log(n)\log(t))$ variance, $O(n)$ space and $O(n \log n)$ pre-processing time for the nearly-optimal bounded-input algorithm of Henzinger et al. Empirically, we find that our algorithm's performance is also comparable to theirs in absolute terms: with optimized parameters, our variance is less than $2\times$ theirs for $t$ as large as $2^{32}$.

cs.CR · cs.DS · cs.LG

版本更新 · v2arXiv:2604.06277

Weakly Supervised Distillation of Hallucination Signals into Transformer Representations

Shoaib Sadiq Salehmohamed, Jinal Prashant Thakkar, Hansika Aredla, Shaik Mohammed Omar, Naveen Bera, Pulijala Sai Nikhila, Kondaguduru Abhiram, Shaik Gayaz Ali 等 9 位作者

原文摘要

Existing hallucination detection methods for large language models (LLMs) rely on external verification at inference time, requiring gold answers, retrieval systems, or auxiliary judge models. We ask whether this external supervision can instead be distilled into the model's own representations during training, enabling hallucination detection from internal activations alone at inference time. We introduce a weak supervision framework that combines three complementary grounding signals: substring matching, sentence embedding similarity, and an LLM as a judge verdict to label generated responses as grounded or hallucinated without human annotation. Using this framework, we construct a 15000-sample dataset from SQuAD v2 (10500 train/development samples and a separate 5000-sample test set), where each example pairs a LLaMA-2-7B generated answer with its full per-layer hidden states and structured hallucination labels. We then train five probing classifiers: ProbeMLP (M0), LayerWiseMLP (M1), CrossLayerTransformer (M2), HierarchicalTransformer (M3), and CrossLayerAttentionTransformerV2 (M4), directly on these hidden states, treating external grounding signals as training-time supervision only. Our central hypothesis is that hallucination detection signals can be distilled into transformer representations, enabling internal detection without any external verification at inference time. Results support this hypothesis. Transformer-based probes achieve the strongest discrimination, with M2 performing best on 5-fold average AUC/F1, and M3 performing best on both single-fold validation and held-out test evaluation. We also benchmark inference efficiency: probe latency ranges from 0.15 to 5.62 ms (batched) and 1.55 to 6.66 ms (single sample), while end-to-end generation plus probe throughput remains approximately 0.231 queries per second, indicating negligible practical overhead.

cs.AI · cs.CL · cs.LG

新论文 · v1arXiv:2611.11529

Data Manipulation Attacks Against Machine Learning Integrated Cyber-Physical Systems

Soyon Choi, Scott Alfeld, Meiyi Ma

原文摘要

There is a well-established body of work demonstrating that individual machine learning (ML) algorithms can be manipulated or deceived. However, in many real-world systems, multiple ML algorithms are composed to enable automated decision-making, particularly in Cyber-Physical Systems (CPS). Often such systems use a composition of vision, planning, and other ML models to affect their internal state and ultimate take actions in the physical world. While there is a huge literature on the (in)security of ML algorithms, it is not immediately clear how an attacker might leverage their ability to trick individual learners when attacking the composition as a whole. We develop methods for creating an attack policy (a mapping from state to attack) for a data-manipulating attacker affecting a CPS or other state-machine-based AI compositions. We describe our framework, a basic threat model and corresponding methods for computing an optimal attack policy, as well as planned extensions. The goal of this work-in-progress paper is to stimulate conversation and forge new collaborations regarding the security of compositional learning.

cs.LG

版本更新 · v2arXiv:2610.12289

TestPrism: Rethinking Test Evaluation Beyond a Single Reference

Han Li, Lingxiang Hu, Jiacheng Huang, Ziqian Jiang, Jingkai Luo, Wei Gao, Yunfan Tan, Zun Wang 等 9 位作者

原文摘要

Large language model (LLM) coding agents have advanced test generation across diverse programming tasks. However, the common practice of evaluating tests against a single reference solution overlooks alternative valid implementations and can overstate test quality. We introduce TestPrism, comprising 300 test tasks from 17 sources and 3000 candidate implementations, evenly split between valid and invalid solutions. Its primary metric, Joint Success Function, requires the generated tests to fail on the initial program state, accept every valid candidate, and reject every invalid candidate. Across fourteen baseline coding agent configurations, Joint Success Function reaches only 28.00%, whereas single reference success reaches 59.67%. Our analysis reveals missed behaviors, unsupported assertions, and faulty test construction. To address these weaknesses, we introduce TestHelix, which combines heterogeneous synthesis of test and repair pairs with peer cross validation and recursive self improvement (RSI). Across two models, TestHelix improves Joint Success Function by 8.67 to 9.00 percentage points over the native harness comparators in the TestHelix evaluation

cs.SE

版本更新 · v4arXiv:2505.19912

APE: Selective Fine-tuning with Acceptance Criteria for Language Model Adaptation

Javier Marín

原文摘要

We present Adjacent Possible Exploration (APE), a selective fine-tuning method for adapting large language models that systematically explores parameter modifications while maintaining model stability. Inspired by evolutionary optimization principles, APE evaluates multiple candidate parameter updates through fine-tuning on small data subsets and accepts only those exceeding a performance threshold. Unlike standard fine-tuning that follows single gradient directions, APE implements a filtered selection process that prevents destabilizing parameter changes while enabling systematic improvement. Our method achieves 33.9\% BLEU improvement and 36.2\% perplexity reduction on news summarization tasks while using minimal computational resources. The approach provides a practical framework for controlled model adaptation that balances performance gains with representational stability.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.11526

A Stability Analysis of AdamW: Unstable Equilibria and Non-Convergence

Baiyu Su, Lizhang Chen, Qiang Liu

原文摘要

AdamW is ubiquitous in deep learning, yet its behavior remains poorly understood. We analyze its dynamics through the lens of dynamical systems and show that AdamW admits an implicit fixed-point objective: its fixed points coincide with the stationary points of a constrained and regularized optimization problem. However, not all of these fixed points are stable under AdamW's dynamics, and stability depends sensitively on curvature, weight decay, and momentum parameters. Even in one-dimensional settings, equilibria may be unstable, and trajectories can exhibit persistent oscillations rather than convergence to them. We further extend the analysis to higher dimensions, deriving sufficient conditions for local stability of the continuous-time dynamics, and propose a convergent modification that preserves AdamW's fixed points. These results clarify what optimization problem AdamW is associated with, when its convergence can be expected, and how its dynamics could inspire more reliable optimizers.

cs.LG · math.OC · stat.ML

新论文 · v1arXiv:2611.11525

Boundary Mass in Feature Space as a Label-Budget Diagnostic for Representations

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

原文摘要

Low-label transfer is often decided by a practical question: after a representation looks useful, how many labels are still needed to learn the downstream boundary? Transferability and probe-quality scores characterize global alignment, likelihood, margins, and calibration. Boundary-near target mass complements these diagnostics by measuring the ambiguous regions where scarce labels are most costly. We propose effective boundary complexity, a label-budget diagnostic for frozen representations and linear probes. It measures mass in a thin strip around the learned boundary and normalizes it by class balance, so dense ambiguous regions and predicted-class imbalance are both counted. We prove that, under a smooth representation map, density change and tangential surface change cancel at first order, leaving boundary-normal stretch or compression as the governing factor. This yields a cached-feature estimator that fits a probe on one split and counts independent held-out features near its boundary. Experiments support the diagnostic at three levels. Controlled transformations match the predicted law across 74 maps with 0.38% median relative error. On real frozen features, normal-direction changes move both measured boundary mass and label demand, while tangent controls preserve the first-order coefficient. A differentiable surrogate reduces measured complexity by 16-36% at matched top-1 utility. Finally, CIFAR-100 and Tiny-ImageNet experiments support boundary mass as a second-stage diagnostic for near-tied representations. Five-way experiments show strong OvR-interface backbone-ranking agreement (Spearman 0.872), while acquisition experiments show high correlations between initial boundary mass and realized label budgets across three sampling policies (0.779-0.830). Code is available at https://github.com/qqaqq-code/boundary_mass.

cs.LG · cs.AI

版本更新 · v2arXiv:2401.13191

Towards Multi-domain Face Landmark Detection with Synthetic Data from Diffusion model

Yuanming Li, Gwantae Kim, Jeong-gi Kwak, Bon-hwa Ku, Hanseok Ko

原文摘要

Recently, deep learning-based facial landmark detection for in-the-wild faces has achieved significant improvement. However, there are still challenges in face landmark detection in other domains (e.g. cartoon, caricature, etc). This is due to the scarcity of extensively annotated training data. To tackle this concern, we design a two-stage training approach that effectively leverages limited datasets and the pre-trained diffusion model to obtain aligned pairs of landmarks and face in multiple domains. In the first stage, we train a landmark-conditioned face generation model on a large dataset of real faces. In the second stage, we fine-tune the above model on a small dataset of image-landmark pairs with text prompts for controlling the domain. Our new designs enable our method to generate high-quality synthetic paired datasets from multiple domains while preserving the alignment between landmarks and facial features. Finally, we fine-tuned a pre-trained face landmark detection model on the synthetic dataset to achieve multi-domain face landmark detection. Our qualitative and quantitative results demonstrate that our method outperforms existing methods on multi-domain face landmark detection.

cs.CV

新论文 · v1arXiv:2611.11522

Music Inpainting With Similarity-Guided Latent Diffusion

Sean Turland, Eloi Moliner, Vesa Välimäki

原文摘要

We propose a music inpainting technique combining similarity matching and a diffusion-based generative model to replace large segments of audio recordings with suitable alternatives. The proposed hybrid method, Similarity-Guided Diffusion Posterior Sampling (SimDPS), retrieves candidate replacement segments from the wider signal based on contextual similarity and uses them to guide the diffusion process toward coherent reconstructions. Being model-agnostic, the framework is compatible with any pre-trained diffusion model operating in either the waveform or latent domain. Objective evaluations demonstrate that SimDPS outperforms widely adopted RePaint and DPS methods across acoustic features and neural-embedding distance metrics. Subjective evaluations on multi-genre music datasets confirm that SimDPS enhances perceptual plausibility compared to previous diffusion-based methods and demonstrates robustness to poorly matched similarity candidates. These findings pave the way for improved music restoration and context-aware audio editing.

cs.SD · cs.LG · eess.AS

新论文 · v1arXiv:2611.11521

K-PWM: Control-Oriented Structured Bilinear World Models under Partial Observation

Santosh Mohan Rajkumar, Sriram Narayanan, Samuel E. Otto, Debdipta Goswami

原文摘要

World models learn internal simulators of environment dynamics for prediction, planning, and control, yet learning them from limited, noisy, and partial observations remains challenging. While much progress has focused on pixel-based observations, many physical systems instead provide vector-valued measurements that reveal only part of the underlying state. We introduce K-PWM, a probabilistic world model for prediction and control under partial observation. K-PWM couples Koopman-bilinear latent dynamics with a nonlinear observation decoder and performs probabilistic latent-state inference through the learned model via extended Kalman filtering and smoothing, without a separately trained inference network. Its bilinear structure enables exact propagation of the maintained Gaussian belief through the latent transition and efficient rollouts for uncertainty-aware model predictive path integral (MPPI) control. For training K-PWM, we present a generalized expectation conditional maximization (GECM) procedure with closed-form updates for the generators, noise covariances, and latent prior, complemented by gradient-based decoder learning. A Koopman-consistent initialization selects the latent dimension from data and initializes the action-dependent dynamics. With limited training data, K-PWM achieves lower long-horizon prediction error than the evaluated baselines on partially observed Gymnasium MuJoCo systems, while substantially reducing training time relative to the strongest prediction baseline. On high-dimensional MuJoCo Menagerie robots, it achieves competitive tracking accuracy with low per-step computation times, demonstrating potential for real-time deployment. These results support structured probabilistic dynamics as a data- and compute-efficient approach to world modeling and control under partial observation.

cs.RO · cs.AI · cs.LG · eess.SY · math.DS

新论文 · v1arXiv:2611.11520

Hydra: Towards Transferable Multi-Task Learning on Temporal Graphs

Kiarash Shamsi, Farimah Poursafaei, Tran Gia Bao Ngo, Reihaneh Rabbany, Baris Coskunuzer, Guillaume Rabusseau, Michael M. Bronstein, Shenyang Huang 等 9 位作者

原文摘要

Real-world evolving networks are naturally modeled as temporal graphs (TGs), where capturing temporal dynamics is essential for predicting future graph properties that support downstream decision-making. Existing temporal graph methods have been developed primarily for single-task prediction, and little is known about their generalization across tasks or transfer to unseen networks. This leaves the challenge of multi-task graph property prediction in TGs largely open. We address this challenge by introducing Hydra, a novel architecture that integrates local connectivity features from temporal GNNs with a spectral learning module that captures global connectivity patterns. This design enables joint learning of local and global information under a multi-task objective. In multi-task classification, Hydra achieves an 8.9% relative gain in AUC over the strongest competitor. In multi-task regression, Hydra achieves competitive results in all three tasks, while obtaining the best results in two tasks with a 8.2% relative gain in MAE compared to the strongest baseline. Moreover, Hydra delivers these gains with a 22% reduction in training time compared to temporal transfer models. These results provide the first systematic evidence that multi-task transferable learning on temporal graphs is effective. By delivering consistent top-ranked performance, Hydra highlights multi-task training on temporal graphs as a promising direction toward adaptable foundation models for temporal graphs.

cs.LG · cs.AI

新论文 · v1arXiv:2611.11519

From MNH to the Friedman--Komogortsev Method---which changes improve eye-movement classification?

Lee Friedman, Mehedi Hasan Raju, Oleg Komogortsev

原文摘要

The Friedman-Komogortsev Method (FKM) is a rule-based classifier of fixations, saccades, and post-saccadic oscillations in 1000-Hz recordings, developed from the modified Nyström-Holmqvist algorithm (MNH) and used in several recent studies. FKM changes MNH in many ways at once: it centres MNH's Savitzky-Golay filter, adds a position-based test for whether an apparent post-saccadic oscillation continues the saccade, adds rejection tests, and changes several thresholds. We asked which of these changes alter agreement with human classification. Both algorithms were scored against twenty 26-s GazeBase reading recordings hand-classified by three trained raters, and each change was reverted in turn. FKM agreed better than MNH as MNH was used: saccade Cohen's kappa rose on all 20 recordings (median +0.034), and saccade boundaries moved from 2-3 ms late to the raters' sample. One change produced this: centring the filter, which removes a three-sample delay. Reinstating the delay returned FKM's saccade kappa to MNH's value, and centring MNH's filter alone made MNH match FKM on saccades and exceed it on post-saccadic events. The position-based test changed how saccades with several velocity peaks were divided, splitting 8% of rater-labelled "wavy" saccades into a saccade and a spurious oscillation against 39% for delay-corrected MNH. FKM's higher post-saccadic velocity floor traded recall for precision. The remaining changes had no measurable effect on these recordings. FKM's improvement over MNH is therefore a filter-alignment correction that MNH users can apply directly, plus a segmentation rule for multi-peak saccades.

q-bio.NC · cs.HC

新论文 · v1arXiv:2611.11517

InfCoiL: Coordinated Planner-Controller Learning for Closed-Loop Physics-Based Human-Object Interaction

Yude Zou, Yifei Yao, Yiwei Hao, Hanqing Wang, Xing Gao

原文摘要

Text- and goal-conditioned human-object interaction (HOI) provides an intuitive interface for specifying high-level intents in physics-based character animation. Mapping such intents to physically plausible, contact-rich behaviors naturally calls for a generative planner and a physics-based controller, yet existing planner-controller systems either execute plans open-loop or close the loop while keeping the planner fixed, so execution feedback never reshapes what is planned. We present InfCoiL, a closed-loop framework for physics-based HOI from coordinated planner-controller learning. InfCoiL couples a morphology-aware flow-matching planner with a unified multi-morphology controller through a shared morphology interface in an autoregressive plan-and-execute loop, and alternately fine-tunes both components with reinforcement learning from the same physical rollouts. To optimize the flow-matching planner online, we introduce Semantic Flow Policy Optimization (SFPO), which stochasticizes selected denoising transitions with channel-wise semantic covariance over heterogeneous HOI features and assigns group-relative, sequence-level rewards from complete closed-loop executions. On held-out OMOMO sequences, InfCoiL outperforms open-loop hierarchical baselines in goal-reaching task success and physical tracking accuracy. Project page: https://infcoil.github.io/

cs.CV · cs.LG

版本更新 · v2arXiv:2602.04937

Linear Model Merging Unlocks Simple and Scalable Multimodal Data Mixture Optimization

Davide Berasi, Matteo Farina, Massimiliano Mancini, Elisa Ricci

原文摘要

Selecting the best data mixture is critical for successful Supervised Fine-Tuning (SFT) of Multimodal Large Language Models. However, determining the optimal mixture weights across multiple domain-specific datasets remains a significant bottleneck due to the combinatorial search space and the high cost associated with even a single training run. This is the so-called Data Mixture Optimization (DMO) problem. On the other hand, model merging unifies domain-specific experts through parameter interpolation. This strategy is efficient, as it requires only a single training run per domain, yet it often leads to suboptimal models. In this work, we take the best of both worlds, studying model merging as an efficient strategy for estimating the performance of different data mixtures. We train domain-specific multimodal experts and evaluate their weighted parameter-space combinations to estimate the efficacy of corresponding data mixtures. We conduct extensive experiments on 14 multimodal benchmarks, and empirically demonstrate that the merged proxy models consistently select near-optimal mixtures. This decouples the search for optimal mixtures from the resource-intensive training process, thereby providing a scalable and efficient strategy for navigating the complex landscape of mixture weights. Code is publicly available at https://github.com/BerasiDavide/mLLMs_merging_4_DMO.

cs.LG · cs.AI · cs.CL

版本更新 · v3arXiv:2604.07923

Stitch4D: 4D Urban Reconstruction from Sparse Multi-Location Observations via Spatio-Temporal Interpolation

Hina Kogure, Kei Katsumata, Taiki Miyanishi, Komei Sugiura

原文摘要

Dynamic urban environments are often captured by cameras placed at spatially separated locations with little or no view overlap. However, most existing 4D reconstruction methods assume densely overlapping views and struggle under sparse multi-location observations, producing unstable reconstructions in unobserved intermediate regions. To address this practical yet underexplored setting, we propose Stitch4D, a unified 4D reconstruction framework that compensates for missing spatial coverage in sparsely observed urban environments. Stitch4D synthesizes intermediate \emph{bridge views} between distant camera locations and jointly optimizes real and synthesized observations in a unified coordinate frame with inter-location consistency constraints. By recovering intermediate spatial coverage before optimization, Stitch4D mitigates geometric collapse and improves reconstruction stability in sparse regions. To evaluate this setting, we introduce Urban Sparse 4D (U-S4D), a controlled CARLA-based benchmark for free-viewpoint reconstruction under sparse multi-location configurations. Experiments on U-S4D show that Stitch4D consistently outperforms representative 4D reconstruction baselines in image-quality metrics. These results suggest that recovering intermediate spatial coverage is an effective strategy for stabilizing 4D reconstruction in sparse urban environments. The project page is provided in https://stitch4d-project-page.vercel.app/.

cs.CV

版本更新 · v2arXiv:2302.12670

Personalized Pricing with Invalid Instrumental Variables: Identification, Estimation, and Policy Learning

Rui Miao, Zhengling Qi, Cong Shi, Lin Lin

原文摘要

Pricing based on individual customer characteristics is widely used to maximize sellers' revenues. We study how a firm can learn such a personalized pricing policy from historical transactions when unobserved factors affect both the prices it charged and the revenue it earned. Standard instrumental variable (IV) methods either focus on a discrete treatment space or require that the IV affect revenue only through price. This exclusion restriction is difficult to justify in pricing, where the most natural candidate IVs are price-setting levers that may also affect revenue directly. We propose PRINT (Personalized pRicing using Invalid iNsTrumental variables), a policy learning method for continuous prices that uses such candidate IVs. Relying on structural models of revenue and price, we show that the optimal pricing policy can be identified with invalid IV that shifts higher moments of historical prices, when the unobserved factors that drive customers' price sensitivity are separated from those used by the historical pricing rule. Based on this identification, which leads to conditional moment restrictions with generalized residual functions, we construct an adversarial min-max estimator that accommodates flexible machine learning models and establish regret bounds for the learned pricing policy. Simulation studies and an application to a U.S. online auto lender illustrate the method's performance. Our results give a principled way to learn personalized prices from observational data using imperfect price-setting levers, and clarify when doing so is justified.

stat.ME · cs.LG · econ.EM · stat.ML

新论文 · v1arXiv:2611.11516

Breaking survivorship bias in driving world models

Zihao Sheng, Zilin Huang, Yuhao Luo, Zhengyang Wan, Boyue Wang, Soyoung Ahn, David A. Noyce, Sikai Chen

原文摘要

World models allow autonomous vehicles to imagine the consequences of candidate actions before execution, but their reliability depends on the futures represented in their training data. We identify a survivorship bias in driving world models: because the futures recorded in prevailing driving datasets almost always end safely, existing models often imagine benign outcomes even when a collision is imminent. We introduce a neutral world model (NWM) that learns both routine and safety-critical futures from a mixture of routine driving and real accident videos with a bounding-box-guided objective. The NWM more faithfully anticipates imminent collisions, while maintaining a low false collision rate and competitive generation quality in routine driving. We further use the model as an imagination engine for decision-making. Given multiple ego trajectories, the NWM predicts the future associated with each candidate, and a reward function evaluates these imagined outcomes before selecting an action. On challenging NAVSIM scenarios, this procedure selects the better trajectory more reliably than the baseline world model. Zero-shot experiments with real-world pedestrian dummy crossing data collected using an autonomous research vehicle further demonstrate transfer to unseen sensor observations. These results show that correcting survivorship bias enables world models to more faithfully reason about futures and support safety-critical decision-making in autonomous driving.

cs.RO

版本更新 · v3arXiv:2609.04241

VocalCoachBench: Benchmarking Audio-Language Models on Expert Feedback for Singing

Hayeon Bang, Hounsu Kim, Wonil Kim, Juhan Nam

原文摘要

Recent audio-language models are increasingly evaluated on recognizing, describing, and reasoning about audio, but expert-facing applications require a different capability: producing feedback that identifies problems and suggests corrective actions grounded in the input. We introduce VocalCoachBench, a benchmark for evaluating audio-language models on expert vocal coaching feedback for singing. VocalCoachBench contains 515 recordings annotated by 18 professional vocal trainers, yielding 1,064 expert audio-level reviews and 12,216 atomic coaching claims. It comprises a same-song subset for controlled comparison and a diverse-song subset for segment-grounded feedback across varied songs and recording conditions. To accommodate the open-ended nature of expert feedback, VocalCoachBench separates deterministic structured targets from claim-based assessment of free-form diagnosis and corrective guidance. Human annotation analysis shows that expert agreement varies strongly with label granularity, motivating hierarchical structured metrics and claim-based evaluation of open-ended feedback. Experiments with 12 recent audio-language models reveal a persistent gap: while the strongest models can compare performances and identify broad issue domains in free-form feedback, the best Top-3 macro-F1 only slightly exceeds a label-prior baseline, and strict diagnosis hit rates remain below 8%. To our knowledge, VocalCoachBench provides the first public testbed for evaluating audio-grounded expert feedback for singing, moving audio-language evaluation beyond description toward analytic feedback. Data, code, and audio examples are available at https://hayeonbang.github.io/VocalCoachBench.

cs.SD · eess.AS