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

+1112新增论文
+875其中 AI 相关
+133已有论文修订
+97新增 AI 导读

arXiv 最近完成扫描:2026-11-20 03:37 · 全站最近入库:2026-11-19 19:09各信源进度 →

统计口径与数据新鲜度

统计 2026-11-19 04:06 — 2026-11-20 04:06(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:2026-11-20 03:37最近发现新增版本:2026-11-19 19:09库内论文最新更新时间:2026-11-19 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。

27,982 已收录论文23,249 AI 相关8,559 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 30 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-20 03:35(北京时间)
新论文 · v1arXiv:2611.23753

Interpreting Timestamp Permutations in Dynamic Link Prediction

Ilia Karpov, Kirill Koloskov, Valeriy Kuleshov, Dmitry Ilvovsky

原文摘要

Timestamp permutations are used to test whether dynamic link predictors exploit temporal information, but a performance gap can reflect changes in training, evaluation history, or both. We separate these effects with a crossed diagnostic and measure input exposure: whether altered histories actually change the predictor's inputs. On MOOC, TCL trained with binary cross-entropy gains approximately eight pairwise-accuracy points from original-chronology training under either evaluation history, while changing history at fixed weights gives only 0.03 points. The small history contrast reflects unchanged inputs for 68.37% of queries, stable rankings for most others, and cancellation of gains and losses. Under original evaluation history, suppressing explicit time encoding reduces the training contrast from 8.03 to 3.64 points, a paired reduction of 4.39 points (95% interval [3.68, 5.09]) across six seeds. DyGFormer shows corresponding training and history contrasts of 22.34 and 0.38 points across two seeds, despite greater exposure. Partial history permutations further separate nominal intervention strength, input changes, and aggregate accuracy. These comparisons turn a single permutation gap into interpretable statements about training conditions and model responses, while leaving the internal mechanisms of temporal learning open.

cs.LG · cs.AI · cs.LO · cs.SI

新论文 · v1arXiv:2611.22434

Encoding, Decodability, and Use: A Five-Layer Framework for Benchmark Measurement Audits with Evidence from UAV Intent Recognition

Zhirong Su, Tian Zhang

原文摘要

Benchmark accuracy is routinely read as evidence that a model understands the construct the benchmark measures. We propose and instantiate on AeroIntent a five-layer framework that decomposes that reading into observation, representation, decoding, use, and behaviour. The framework separates encoding, decodability, and use as distinct levels. In the instantiation, the observation layer is unreliable on 7 of the 20 Drone programs and floors the direct visual correlate of radial velocity at a 0.26\% scale change. The representation layer exposes two task-designed RGB features with nominal signals. RAFT flow radial bias reaches $r=+0.257$ ($p=0.030$), and bounding-box optical-flow divergence reaches $r=+0.203$ ($p=0.029$). No multi-feature readout gives a fold-mean test correlation above the cluster minimum detectable effect (MDE), and the largest positive global out-of-sample correlation is $r=+0.080$. The decoding layer sets the MDE at $|r|=0.599$ at the multi-feature Bonferroni threshold ($α=0.00714$, power 0.80), and the observed $r=+0.080$ has power 0.00. Two injection experiments establish the MDE's condition from both sides: aligned injections are detected at or above the simulated power, and orthogonal injections are not detected. The use layer finds accuracy consistent with program memory (program-identity 0.937/0.597/0.210) and multi-region static cues. The behaviour layer is consistent with accuracy not requiring a radial-velocity readout (appearance 0.6552 vs bounding-box trajectory 0.4080, cluster-level paired $Δ=+0.2471$, $p=0.0132$).

cs.LG · cs.AI

新论文 · v1arXiv:2611.22228

What Does Weight-Space Programming Add to Diffusion Editing? Program Construction versus Score-Guided Refinement

Zitian Gao, Jingyi Xu, Ziqi Han, Chengkai Lei, Ye Liu, Liangliang Shi

原文摘要

Weight-space programming (WSP) lets a language model construct executable weight updates for another network. We study diffusion concept editing through an interface in which programs define training contexts, preservation constraints and update directions, while a shared numerical fitter selects their coefficients. We compare generated programs with published editors, a shared-panel hand-written baseline and budget-matched search over conventional edits on 24 TIMED test edits for Stable Diffusion 1.4. The preregistered rewrite pipeline shows no advantage over the strong baseline. In an exploratory comparison, one-shot Astra programs improve fresh-scene VLM joint success, which requires subject presence together with whole-image target matching, by +10.0 points over the shared-panel baseline, while CLIP relative match shows no detectable change. These programs organise adjustable directions around protected text-state subspaces. Probing and rewriting the Astra programs then raises CLIP by +6.2 points while joint success changes by -9.7. On a fixed 144-image sample, five additional evaluators give point estimates in the same directions under whole-image target matching; checks of subject-local change give less decisive differences. Together, these results identify useful behaviour in complete generated programs and show that score-guided revision need not preserve it: program construction and revision should be evaluated as distinct stages of a WSP pipeline.

cs.LG · cs.CV

新论文 · v1arXiv:2611.20696

Conic Structure of the Cross-Entropy Loss at the Final Layer and Its Absence in Rectifier Backbones

Omer Nezih Gerek

原文摘要

Slices of the loss along lines and planes in parameter space are widely used to visualize deep networks. They often show a single basin with nearly straight walls, which raises the question of whether such a loss is a cone rounded at its apex, or a lower envelope of several cones that meet along creases. We answer this question for softmax cross-entropy at two levels of a classifier. Over the weights of an affine final layer, the loss equals a polyhedral cone plus an offset that decreases monotonically and exponentially along every ray, towards a limit in closed form. The distance at which the asymptotic wall reaches the height of the starting point, which we call the core radius, equals the mean surprisal of the classes that the direction favours, divided by the slope of the wall. Over the parameters of a rectifier backbone, the loss along a line has a kink wherever the activation pattern of a sample changes. Using the dimension theory of semialgebraic sets, we show that, along almost every line, the samples that switch at one point share an irreducible factor of their pre-activation polynomials. Under a coprimality condition, each kink therefore carries the share of one sample among $n$, so creases comparable to the slope do not form. Measurements on ResNet-18 and ResNet-56 trained on CIFAR-10 illustrate both results, with a test that bounds the size of any crease that sampled loss values can hide.

cs.LG · eess.SP

版本更新 · v2arXiv:2611.06044

Do Search-Augmented Recommenders Use the Latest Search Episode?

Jiandong Ding

原文摘要

When a user returns to a recommendation feed after searching, does the latest search episode help rank the next item? Aggregate ranking gains leave open whether search-augmented recommenders use the latest search at prediction time or benefit from broader search state and representation learning. FocalSwap is a counterfactual audit that masks or replaces the latest episode or the full search history while holding the target, candidates, recommendation history, and trained scorer fixed. A planted-signal test and GenSAR's native search-item prediction task verify that FocalSwap detects known search dependence. In proxy analyses, KuaiSAR's routing gain falls below the shuffled threshold, and Qilin's clearest matched-search advantage appears at high taxonomy overlap. The primary comparisons for SESRec and UniSAR do not establish a positive ranking gain from the latest search episode alone. For the full search history, the estimated contribution depends on the control. On 66,139 shared instances across three UniSAR checkpoints, neutral masking consistently lowers mean reciprocal rank (RR). Replacing the history with observed searches matched on category and activity profile changes 44.1 to 44.8 percent of target ranks, yet the original history has little average advantage. In the primary assignment, matched replacements yield mean RR 0.002770 to 0.006604 higher than neutral masking across checkpoints. The adjusted intervals for this difference lie above zero under the primary assignment and all five alternative assignments per checkpoint. Search audits should specify which inputs changed and what replaced them, and report both how often ranks move and how much the original search helps.

cs.IR · cs.LG

新论文 · v1arXiv:2611.19741

Where the Error Lives: Separating Lid-Edge and Body-Induced Error in Neural Pressure Prediction for 3D Cavity Flow

Zirun Zhou, Zhi-Qian Zhang, Yixuan Hong, Guang Liu, Zhigang Liu, Lili Zhang

原文摘要

Can a pressure surrogate score much better globally without better resolving the pressure around immersed bodies? We test this in a three-dimensional lid-driven cavity with an unregularized lid and centered pressure at native $128^3$ resolution. On 60 cases from 12 unseen geometries, with predictions sealed before the CFD truth existed, adding a body-free, Reynolds-conditioned reference correction to the same frozen $64^3$ surrogate lifted to $128^3$ lowers the mean full-volume relative $L_2$ error from 20.6% to 7.5%. In the post hoc regional analysis, the lid-edge junction occupies 1.5% of the fluid but holds 69% of the truth pressure energy and 95% of the uncorrected surrogate's squared error. Interior error changes by just +0.16% relative, and object-local peak error and recall change little on average. Retrospective ablations locate the learned term's contribution in the interior and near the bodies; rebasing on the reference lowers global error further with little change in cohort-mean near-body metrics. Reference-relative body-response error remains about 25% in the interior near-body band. Grid-refinement, lid-regularization and convergence checks delimit the numerical scope. In this benchmark, a large global gain chiefly reflects correction of a shared boundary field, not better resolution of the body response. We recommend reporting regional truth-energy and model-error shares alongside near-body and reference-relative body-response errors.

physics.flu-dyn · cs.LG

版本更新 · v2arXiv:2603.00496

ESENSC: A Polynomial-Time Axiomatic Alternative to SHAP

Kazuhiro Hiraki, Shinichi Ishihara, Takumi Kongo, Junnosuke Shino

原文摘要

We propose ESENSC, a computationally efficient and theoretically grounded alternative to SHAP. ESENSC is a deterministic, closed-form attribution rule that requires only linearly many coalition-value evaluations, satisfies the null-player property, and admits an axiomatic characterization based on efficiency, a restricted form of differential marginality, and explicit computational constraints.Empirically, on tabular prediction tasks using neural network and XGBoost models, ESENSC shows close agreement with exact SHAP in both attribution magnitude and feature rankings while requiring substantially less computation. Under the evaluated settings, it compares favorably with widely used sampling-based SHAP approximations. While exact SHAP requires exponentially many coalition evaluations, ESENSC scales linearly with the number of features. These results show that ESENSC offers a favorable trade-off between agreement with exact SHAP and computational cost, providing a practical and theoretically grounded alternative for feature attribution in high-dimensional applications.

cs.LG · cs.AI

新论文 · v1arXiv:2611.15829

Training Calibrated Conditional Generative Models with a Sample-Only Coverage Loss

Benjamin Remy, Raphael Rossellini, Chihway Chang, Rebecca Willett

原文摘要

Conditional generative models have become the tool of choice for sampling posteriors in scientific inverse problems, where we seek to sample images or signals conditioned on a set of indirect and noisy measurements. Yet they are often overconfident, which is a critical failure in scientific applications where accurate uncertainty quantification is essential. Existing regularizers fall short in complementary ways: differentiable coverage objectives require the model's density and do not identify the posterior, while strictly proper scoring rules such as the energy score need only samples but do not directly enforce coverage. We close this gap by introducing a differentiable loss function based on the expected coverage probability, estimated from samples alone. It applies to any conditional generative model, with no need for tractable densities or architectural changes. We prove that it is a strict divergence, vanishing if and only if the model equals the true posterior. On simulation-based inference benchmarks, it matches a density-based coverage loss without evaluating the density. On high-dimensional imaging inverse problems, it yields better-calibrated posteriors than scoring-rule regularizers, with little change in the reconstruction quality of point estimates.

cs.LG · stat.ME · stat.ML

新论文 · v1arXiv:2611.15102

One Fabrication Is Enough: A Fake News Article Changes the Forecasts of LLM Forecasting Systems

Zeyu Zhang, Bradly C. Stadie

原文摘要

LLMs are increasingly used for forecasting, and the best forecasting systems retrieve news as evidence to strengthen their predictions. This also opens them to attack. We study the weakest attacker of this kind, who knows only the forecasting question and publishes fabricated news articles, with no access to the model, the retriever or its queries. We evaluate how far this attack reaches on stock questions and real prediction-market questions. Once the article is in the model's context, it reverses 74% to 82% of the forecasts that opposed it, on every frontier model we tested. The model cannot verify the fabrication and follows it anyway. Retrieval is the attack's entry point, and the systems built to retrieve well are the most exposed: the architectures of two leading deployed forecasting systems, whose agents search for every related article, retrieve the fabricated one in about 90% of runs. Accuracy cannot reveal this harm: the attacked forecaster is right when the attacker's direction turns out right and wrong otherwise, so its average accuracy barely changes while its forecasts follow the attacker's guess. We also propose a second attacker, who publishes two articles that argue opposite directions to make the forecaster unsure. The model does not hedge; it returns a confident forecast in a direction the attacker no longer controls. We then evaluate the two strongest published retrieval-time defenses, passage filtering and partitioned retrieval. Each stops the one-article attack but is bypassed by a one-prompt change to the attack. Together they stop both bypasses at four articles, but sixteen distorted copies beat them, and they change 29% of clean decisions and discard 39% of the genuine news. Partitioned retrieval also requires control of the retrieval index. Systems that search the open web have no such control and are left with the filter alone. For them, the attack still works.

cs.CR · cs.CL · cs.LG

新论文 · v1arXiv:2611.14974

Who Moved the Frontier? Paying Open Frontier Challenges with Free Identities

Luca Zanolini, Duc V. Le

原文摘要

In an open frontier challenge, anyone, person or AI agent, submits changes to a shared problem, and a change that beats the best measured score becomes the public baseline. How should it pay when identities are free? Rank prizes can be gamed: on a replay of one track of a recent challenge, one solver's steps under two identities would have taken first and second prize. We consider a prediction-market rule, proposed for machine-learning contests, that pays each accepted step for the progress it adds; with exact scores, among rules that see only a step's start and end, only these resist splitting and merging, and extra identities gain nothing. With noisy scores and independent normal runs, we prove that any payment that never takes money back and pays real improvements also pays something, on average, to a change that does nothing. We propose a fee per submission, at least what luck pays per attempt, that removes each attempt's expected profit if verifiers are honest and no submission is noisier than planned; nothing enforces that bound, and under noise the rule does not cap the total paid. If crediting another and faking credit to one's second account look the same, a rule that neither rewards fakes nor punishes honesty pays for credit at most what lying costs. We introduce a royalty on how much a later gain falls without the earlier change, which needs no claims but needs removable changes. For five public challenges, we name the rule our results would support.

cs.GT · cs.AI · cs.LG

新论文 · v1arXiv:2611.13295

Time Series Change Point Detection with Gaussian Processes

Olga Isupova, Danil Kuzin, Fredrik Gustafsson, Lyudmila Mihaylova

原文摘要

This paper proposes a novel general Gaussian process framework for change point detection in time series. A change is defined as an alteration of the hyperparameters of the Gaussian process prior. We formulate the problem in the context of statistical hypothesis testing and demonstrate its performance over both offline and online multiple change points detection. The approach provides theoretical guaranties of its performance. An important advantage of the developed framework is that it does not need a detailed specification of the time series function before and after a change. The hyperparameter values are learned from the data. Comprehensive evaluation of the proposed framework is provided both on synthetic and real data and its efficiency is demonstrated.

stat.ML · cs.LG · math.ST · stat.ME

新论文 · v1arXiv:2611.12871

Vendi 异常分数:高效准确的异常检测

Vendi Anomaly Scores for Efficient and Accurate Anomaly Detection

Amey Pasarkar, Adji Bousso Dieng

AI 速读 基于摘要

作者提出从多样性视角定义异常检测的 Vendi 异常分数(VAS),通过移除样本后数据集多样性的变化来识别异常,并在多组基准嵌入数据集与 ImageNet 实验上报告了有竞争力的结果。

cs.LG

查看原文摘要

Anomaly detection (AD) is a critical for trustworthy machine learning and scientific discovery. Existing methods typically define anomalies through local density, isolation heuristics, or distance from a fitted distribution, making them sensitive to neighborhood scale, partitioning choices, or restrictive distributional assumptions. In this work, we introduce a new paradigm by formulating anomaly detection from a diversity perspective. We propose the $\textit{Vendi Anomaly Score}$ (VAS), which detects anomalies by quantifying how much a sample changes the diversity of the dataset, as measured by the Vendi score, when removed. VAS captures both local redundancy and global data structure without relying on density estimation or partitioning heuristics. Moreover, VAS requires no hyperparameter tuning, instead adapting automatically to the spectral structure of the data. VAS is non-parametric and scales linearly with dataset size. Across 148 benchmark embedding datasets derived from 10 source datasets and 4 embedding architectures, VAS achieves state-of-the-art AD performance. We further validate VAS on large-scale ImageNet experiments, where it remains robust as both contamination rate and dataset scale increase.

新论文 · v1arXiv:2611.12634

How Far Does Second-Order Linear Response Predict Quantization Mismatch? Pre-Registered Measurements on Five Language Models

Dezhen Lu, Zixiang Ni

原文摘要

Serving systems choose a weight format for each model and deployment. A low-cost screen predicts the resulting KL divergence from the unquantized model's Fisher information using $\tfrac12Δ^\top FΔ$, the expansion underlying Hessian-aware and second-order post-training quantizers. We assess the reliability of this screen in GPU batches with channel-wise int8, per-tensor fp8 and group-wise int4 on Qwen3-8B, Qwen3-4B, OLMo-2-7B, Qwen3-32B and Qwen3-30B-A3B. Relative predictions hold across a 16-fold range of perturbation scales, int4 group sizes and rounding modes, while absolute predictions are heavy-tailed. A trimmed average brings 10 of 12 cells on Qwen3-4B and OLMo-2 within 15% of measured KL. The MoE model exhibits nonquadratic response at small perturbations, with int8 KL scaling as $α^{0.63}$. Computing per-sequence gradients in bf16 can cause fourfold prediction errors. We trace these errors to arithmetic in the lower half of the backward pass and eliminate them by using fp32 there. Batch-invariant kernels run at 0.55 times the throughput, while changes in reduction order introduce a floor of 0.17-0.22 nats per sequence. We build a screener with one streamed fp32 Fisher pass, trimmed tensor-block predictions with a band, a flag for aligned tensors and per-layer format allocation. On the unseen Mistral-7B-v0.3 family, all six predictions fall within the band, all three KL budgets are met, and vLLM throughput is predicted within 0.4%.

cs.CL · cs.CR · cs.LG

新论文 · v1arXiv:2611.12225

A Cost That Practice Accuracy Hides: Worked Solutions in Self-Generated Training Data and the Compositional Use of New Rules

Jiantong Xu

原文摘要

A worked solution can make self-generated training targets shorter and cheaper, but accuracy on the practiced operation may hide how well a model can reuse newly learned rules. We study this choice in a 2B-parameter reasoning model learning two invented quantitative rules. The model generates targets from a rule sheet alone or from the sheet plus a program-written worked solution; targets are accepted for final-answer correctness, and the support is removed for fine-tuning and evaluation. Training questions request forward use in one segment; composition requires sequential reuse of the same rule across two segments. After six passes, using a program-checked numerical signature, sheet-only targets lead by 24.3 percentage points (95% item-bootstrap CI 14.6-34.0) on composition in a prospective test. On unseen argument values, observed forward accuracy is identical (87.5% each), but composition accuracy is 58.3% versus 23.6%, a 34.7-point gap (25.7-43.8). An independently sampled second target-bank pair repeats the direction (+14.6 points, 2.1-26.0). Ten passes on the shorter worked-solution targets, approximately matching the domain tokens of six sheet-only passes while changing other training quantities, narrow the gap to 16.3 points (6.3-26.7) in the first bank pair and 6.2 points (-6.3 to 18.8) in the second, where the pre-specified test does not hold. At six passes each, the worked-solution policy uses fewer capture tokens and less update time, yet the trained models use 2.6 times as many generated tokens per composition success under the numerical signature (2.1 times in the second bank pair). Choosing a capture context therefore requires testing how the new rules are reused, alongside the inference cost of those uses, rather than relying on forward accuracy alone.

cs.SE · cs.CL · cs.LG

版本更新 · v2arXiv:2608.19861

PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM Agents

Seongjae Kang, Taehyung Yu, Sung Ju Hwang

原文摘要

Customer-service LLM agents must follow organizational policy when acting on a user's behalf. Compliance failures arise from either forbidden actions, such as granting an ineligible change, or omitted procedural requirements, such as identification or confirmation. Runtime safeguards can intervene on risky actions, but action-local checks do not guide an agent through a multi-step procedure. Workflow-following systems support prescribed process execution, but primarily target workflow completion rather than safeguarding agent behavior. PolicyGuide instead compiles each domain policy into a workflow graph and invokes a proactive verifier at user-turn boundaries. From persisted graph state, the verifier reconciles open requests and returns step-specific remediation along a policy-compliant path. Across the $τ^2$-bench airline, retail, and telecom domains with a GPT 5.4 agent and verifier, PolicyGuide raises mean $\mathrm{Pass}^4$ from $0.42$ to $0.62$, with the largest gain on telecom ($0.19$ to $0.61$), the most workflow-structured domain. The same workflows transfer to Claude Sonnet 4.6 and Gemini 2.5 Pro agents. Complementary evaluations find the lowest observed attack-success rate under adversarial users and the strongest procedural compliance in an independently validated workflow-level audit.

cs.AI · cs.CL · cs.LG

版本更新 · v3arXiv:2606.16933

A Unified Taxonomy and Evaluation Framework for Distributional Shifts in Reinforcement Learning

Ardianto Wibowo, Paulo E Santos, Amer Baghdadi, Matthew Stephenson, Karl Sammut, Jean-Philippe Diguet

原文摘要

Distributional shift in reinforcement learning (RL) is defined inconsistently across the literature, and it is studied alongside out-of-distribution (OOD) generalization and non-stationarity without a clear account of how the three terms are related. Methods are also often described as robust after evaluation under only a few particular changes, leaving the scope of such claims unclear. We develop a unified taxonomy that brings the three terms under one definition of distributional shift. We factorize the joint distribution of a one-step interaction into five components, each produced by its own generating function: initial-state distribution, observation mechanism, policy implementation (distinct from the policy being optimized), transition dynamics, and reward mechanism. A shift is specified by what changes (the changed components) and when it changes (the interaction regime: OOD, non-stationary, or combined), while changes in interaction data and performance are treated as observed effects. We complement the taxonomy with an evaluation framework that measures immediate generalization and subsequent adaptation, summarizing adaptation in a shift adaptation score (SAS). Published benchmark results, controlled Gridworld experiments with DQN, A2C, and PPO, and a mapping of selected studies show that method responses and rankings vary with the shifted component, and that existing evaluations cover only some components. Robustness claims should therefore be bounded by the components and regimes evaluated, since strong performance under one shift does not establish robustness under others. Identifying an unknown component-level source from observed interaction data remains a direction for future research.

cs.LG · cs.AI

新论文 · v1arXiv:2611.10279

Restoring General and Improving Domain-Specific Safety in Fine-Tuned LLMs Without Retraining

Vishnu Kabir Chhabra, Ding Zhu, Zhihui Zhu, Xueru Zhang, Mohammad Mahdi Khalili

原文摘要

Safety-aligned language models are known to suffer degradation in refusal behavior following domain-specific fine-tuning, even when the fine-tuning data is entirely benign. The predominant mitigation strategy, preference optimization on the fine-tuned model, is computationally expensive and requires domain-specific preference data that is often unavailable in practice. We propose SPECTRA, a training-free, weight-level method that restores refusal behavior by transferring the base model's refusal direction into the fine-tuned model's weights via an activation-aware, low-rank update, requiring no gradient-based optimization or domain-specific safety data. The low-rank design restricts the intervention to the subspace most causally active on harmful inputs, leaving benign behavior largely undisturbed by construction. Empirically, SPECTRA substantially improves refusal behavior across a comprehensive safety evaluation, while preserving general coherence, domain-specific knowledge, and producing negligible change in false refusal rates.

cs.AI · cs.LG

新论文 · v1arXiv:2611.08936

Ordered Policy Optimization

Dongkyu Derek Cho, Pan Xu

原文摘要

Proximal Policy Optimization (PPO) stabilizes policy updates by clipping the importance ratio between the new policy and the data-collecting policy. The standard rule uses one fixed clipping range for all samples. While this choice offers simplicity and effectiveness, it fails to distinguish statistically harmful extreme ratio values from large, yet tolerable changes. We study how to adapt the clipping radius to the reliability of the induced importance ratio distribution. Our case study demonstrates that reliable updates require controlling these extreme ratios, rather than merely verifying that individual ratios fall within a fixed interval. Motivated by this observation, we propose Ordered Policy Optimization (\textsc{OPO}). In each minibatch, \textsc{OPO} sorts the importance ratios, selects the extreme ratios that potentially have the largest effect on the update, and assigns them adaptive sample-wise radii instead of using the fixed clipping radius. These radii are derived from a prescribed tail profile, which specifies desired target values for the selected extreme ratios. The resulting objective remains first-order and requires only minibatch sorting and a modified policy loss. Experiments conducted on MuJoCo and Atari benchmarks demonstrate that this simple modification achieves strong performance relative to established baselines.

cs.LG · cs.AI · cs.RO · stat.ML

新论文 · v1arXiv:2611.08568

Personalized health coaching via coachable health age and counterfactual wearable trajectories

Simon A. Lee, Zeinab Esmaeilpour, Menglian Zhou, Marius Guerard, Sohrab Sidhu, Alexandru Solot, Hamid Palangi, Pramod Rudrapatna 等 20 位作者

原文摘要

Personalized health coaching requires moving beyond retrospective wearable summaries toward quantitative, actionable behavioral prescriptions. We introduce Coachable Health Age (CHA), an interpretable health-age metric built from modifiable physiological and behavioral signals passively captured by consumer wearables. Across the WEAR-ME (N=2,780), All of Us (N=22,748), and longitudinal Fitbit (N=1,464) cohorts, CHA aligns with established cardiometabolic health constructs and improves prognostic discrimination for incident cardiovascular and metabolic disease beyond age, sex, and body mass index. To enable prospective behavioral optimization, we developed Sensor World Model (SensorWM), an action-conditioned wearable world model pretrained on over 2.2 million person-weeks of longitudinal sensor data from approximately 100,000 participants. SensorWM accurately forecasts multi-week physiological responses to candidate lifestyle changes ($R^2=0.664$), outperforming generic time-series foundation models while preserving downstream clinical signal for diabetes, hypertension, and other health outcomes. Finally, we integrate CHA and SensorWM into an agentic counterfactual coaching system to synthesize personalized 12-week coaching plans. In blinded paired reviews, certified lifestyle coaches preferred counterfactual-grounded plans over plans grounded in wearable history alone for overall quality (71.4%, P < 0.001) and anticipated adherence (71.0%, P < 0.001), and board-certified clinicians judged 99.2% of counterfactual-grounded plans to meet clinical safety criteria (vs. 97.6% of comparison plans). Automated evaluations across the full cohort (N=2,780) also favored counterfactual-grounded plans. By closing the loop between longitudinal tracking, trajectory prediction, and counterfactual coaching, our framework provides a scalable path to systematically improve cardiometabolic health.

q-bio.QM · cs.LG

版本更新 · v2arXiv:2610.12532

LLM驱动音色生成与插值中的结构与表征能动性

Structural and Representational Agency in LLM-Driven Timbre Generation and Interpolation

Joseph Cameron, Alan Blackwell

AI 速读 基于摘要

在一项LLM驱动音色界面的测量研究中,作者报告用户实际能改变的声音空间与界面所报告的位置可以分离:前者由结构精确刻画,后者存在显著误差。

cs.SD · cs.HC · cs.LG · cs.MM · eess.AS

查看原文摘要

Generative creative interfaces routinely report the user's position within a space of possibilities. Such displays imply that the quantities shown were measured from the artefact; in a growing class of systems they are instead predicted by a language model describing its own output, and the two are indistinguishable on screen. Most creative domains lack an agreed measurement against which such a self-report can be checked; audio admits one. In a measurement study of a working LLM-driven timbre interface, what a user can cause proves separable from what the interface reports. The first is exactly characterisable: the reachable set is a zonotope in parameter space whose dimension increases by one per sound placed, so the space the model may act within is demonstrably the user's own. The second is inaccurate. The displayed position differs from measurement by approximately three times the width of that region, and the error has two independent sources: model-predicted coordinates, whose accuracy varies twofold between the two models evaluated, and the interpolation step, which arises from combining synthesiser parameters and is unaffected by model quality. End to end, 27 of 36 refinement instructions produced no perceptible change, and under the weaker model the facility performed no action while reporting convergence. Effective agency (what a user can cause) and epistemic agency (whether they can know it) are therefore separable and independently engineered: structurally enforced constraints survive model failure, whereas guarantees resting on the model's own report do not. The failure mode that remains is silence presented as success. We close with a short protocol for building and auditing such displays.

版本更新 · v2arXiv:2609.09986

A Sharp Barrier of Consistent Submodular Maximization: Any Improvement over $2-\sqrt{2}$ Entails Exponential Queries or Linear Recourse

Shi Fu, Qixin Zhang, Dacheng Tao

原文摘要

Consistent submodular maximization aims to maintain a high-utility subset of size at most $k$ as new elements arrive one by one, while changing only a limited number of elements per update. Previously, [DFL+25](STOC '25) established an information-theoretic bound of $2/3$ and gave the first polynomial-time randomized algorithm with a $0.51$-approximation guarantee. However, a fundamental question remained open: Is it possible to design an efficient algorithm for consistent submodular maximization that matches the classical offline $(1-1/e)$-approximation? In this paper, we answer this question in the negative by proving that the sharp approximation threshold achievable with polynomially many value queries and worst-case constant recourse is \[β:=2-\sqrt{2}\approx0.5858<1-1/e.\] Specifically, we develop an improved poly-time randomized algorithm, which can attain a $(β-\varepsilon)$-approximation in expectation at each fixed time with $O(\varepsilon^{-2})$ worst-case recourse. The cornerstone of this algorithm is a novel anchored sampling scheme, which uses only value queries on subsets of the elements seen so far to construct a solution that remains competitive under unknown future arrivals. Conversely, we establish a matching lower bound by showing that any fixed improvement over $β$ requires either exponentially many value queries or $Ω(k)$ changes per update. Furthermore, we characterize the impact of curvature by establishing a sharp threshold of $1-(\sqrt{2}-1)\vartheta$ for consistent submodular maximization, where $\vartheta\in[0,1]$ denotes the total curvature. Surprisingly, for weighted coverage objectives, we devise a poly-time randomized algorithm that can match the optimal $(1-1/e-\varepsilon)$-approximation with $O(\varepsilon^{-1})$ worst-case recourse, thereby separating coverage maximization from general submodular maximization in the consistent settings.

cs.DS · cs.LG

新论文 · v1arXiv:2611.03964

Ermakov--Based embeddings as a parameter-free layer for physics-informed neural networks

Roberto Bernal-Jaquez, Roberto Romero Huerta, Alexander Schaum

原文摘要

Evolution equations with time-dependent coefficients force a neural solver to spend capacity on a scale that the equation already determines. We show that a single classical object, the Ermakov-Pinney scale function $ρ$, can be written into the input layer of a physics-informed neural network as a parameter-free embedding. One construction covers three problem families through the two signs of the Pinney constant $c$: linear diffusion ($c=-1$, $ω\equiv0$), diffusion in a time-dependent confining trap ($c=-1$, $ω\neq0$), and the quantum oscillator with time-dependent frequency ($c=+1$). An exact moment identity gives $ρ$ in closed form from the initial data alone, so no auxiliary network or knowledge of the solution is needed. Our central finding is comparative: each family requires a different component of the construction and has a different honest control. In family I, the diffusive-time reparametrisation alone reproduces most of the gain, so that family is a poor test of the idea and shows why an ablation is mandatory. In family II no time reparametrisation exists, the control degrades accuracy, and the amplitude prefactor becomes essential (40.8x). In family III the adiabatic approximation also fails, and the quadratic phase factor is what matters (5.7x). The method does not help when the reduced solution does not converge to a fixed point: its extrapolation advantage then disappears. The reduction itself is the classical scaling-variable change of the self-similar-asymptotics literature; what is new is the moment identity that makes it architectural for time-dependent coefficients, and the comparative anatomy above.

cs.LG

新论文 · v1arXiv:2611.03222

CERBO: Certified Recourse at a Requested Rate for Gradient-Boosted Trees

Farhan Shahriyar Hossain, Taufikur Rahman Fuad, Md. Zarif Bin Hasnat, Md Abrar Jahin, Md Rizwan Parvez

原文摘要

Recourse-aware learning trains models so that people who receive an adverse decision can reverse it with an affordable change. For gradient-boosted decision trees (GBDTs), the existing method, RABIT, exposes a single recourse weight whose effect on the recourse rate is abrupt and data-dependent, and its training bound cannot certify recourse for any rejected individual. We present Certified Recourse Boosting (CERBO), which takes a requested recourse rate and returns a GBDT with certified actions: it tracks exact additive effort along tree paths, trains against committed boxes (a family of bounds that contains RABIT's and certifies a within-budget action for each training sample) and adjusts the weight after every tree until the certified rate meets the request. A calibration step guarantees the rate on new individuals, and the certifier runs on any tree ensemble, including XGBoost and LightGBM, in milliseconds per person without a solver. In RABIT's own experimental protocol, at equal accuracy, CERBO raises the recourse ratio from 0.90 to 0.99 on FICO and from 0.64 to 0.77 on Bail. Under additive costs on six data sets, it raises the area under the accuracy-recourse frontier in all 12 settings, and in simulated deployments it guarantees recourse to 70% of rejected individuals in 11 of 12 settings, against 2 for RABIT.

cs.LG · cs.AI · cs.CY

新论文 · v1arXiv:2611.02437

Evaluating Synthetic Skin Lesion Images for Classifier Training: A Safety-Aware Algorithm

Amindu Kumarasinghe, Vinh Bui, Reza Ghanbarzadeh

原文摘要

Skin cancer classifiers need large sets of annotated images, which are scarce, especially for rare lesion types. AI-generated images can fill the gap, but a generator produces images of widely varying quality, so selecting which images to retain is important. That choice usually rests on a single quality score, most often the Fréchet Inception Distance (FID), which has not been tested against what a classifier needs and cannot tell whether the generator memorizes its training photographs. We test that practice with a tournament of 16 quality metrics. Each metric alone selects a training subset from a pool of 275,472 images generated by a diffusion model (EDM2), and a classifier trained on each subset scores the metric per class. The tournament shows that no single metric serves every class: FID computed on Inception features gives the best classifier for one lesion type and no measurable advantage over random sampling for another, and the generation setting it prefers is also the one closest to the training images. We turn these per-class measurements into a safety-aware evaluation algorithm that weights each metric by its measured value in each class and removes images too close to a real training image. On the held-out ISIC-2018 test set, the algorithm improves Macro-$F_2$, a recall-weighted measure of classification performance, over the FID baseline for a Swin-B classifier by +0.0136 to +0.0164 across four safety thresholds (Bonferroni-corrected $p < 0.02$) and over random sampling by about +0.03, though not for ResNet-50. Tightening the threshold improves a memorization-safety score with no measurable change in utility, but every curated set remains less safe than random sampling. The results show a modest but statistically significant gain on the primary classifier, and offer a practical, safety-aware procedure for evaluating and using synthetic data.

cs.CV · cs.LG

新论文 · v1arXiv:2611.01815

When Does Latent Prediction Error Support Physical Surprise Claims in Video World Models?

Anton Hristov, Aleksandar Slavov, Vasil Vasilev

原文摘要

Prediction errors from video world models are often treated as physical surprise, but they can also reflect appearance, static content, motion, or ordinary prediction difficulty. We present an evaluation framework that separates what frozen representations contain, how they organize appearance and physical differences, and what changes a model's own prediction error. On IntPhys2, a benchmark of physically possible and impossible events, we compare representations from latent prediction, masked pixel reconstruction, and simple visual statistics. Appearance remains easy to decode, while V-JEPA2 exposes modest information about whether events are possible or impossible. Looking at the latent geometry, appearance changes become smaller compared with differences between videos, their directions line up more consistently across scenes, and their overlap with physical changes decreases in deeper layers. Across action-conditioned models, incorrect actions raise prediction error, while appearance changes also produce substantial responses. In Physion++ video clips, repeating a single frame can produce higher error than the original clip, while changing color throughout the clip can lower it. Using Kubric, a simulation and rendering framework, we find that impossible events can produce lower error than possible events. A fixed error threshold also incorrectly flags many possible events as impossible. A model can therefore represent physical information and respond to incorrect actions without its native prediction error reliably measuring physical surprise.

cs.CV · cs.LG

版本更新 · v2arXiv:2610.37255HF 社区推荐

Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces

Jian Mu, Ruibo Shen, Yao Shu

原文摘要

Language model agents are post-trained and deployed through a harness, the prompts, tools, and control logic that present a task to the model and execute its outputs. Harnesses are often revised and judged by how an already trained model performs under them, which assumes that what the model learned under one harness carries over to the next. We ask what a policy learns from its harness and what survives when the harness changes. Under KL-regularized control, we show that the log odds of an optimal policy recover how strongly its harness favors each action over the others, so a trained policy carries the preferences of its harness. Deployed return then equals the ceiling of the harness, the best return it offers, minus the mismatch of the policy, the return lost to its own suboptimal decisions. Training changes only the mismatch, while changing the harness moves both terms. Comparing harnesses with a fixed model therefore mixes what each harness offers with how well that model uses it. In MiniWoB++, post-training makes a model's preferences specific to its training harness. Reversing the order of the observation fields leaves every optimal decision unchanged yet lowers a trained model's success from 85.0% to 19.2%. Conversely, every new harness that a fixed-model comparison rejects but that raises the ceiling outperforms the original agent after further training under it, and ceiling and mismatch measured beforehand predict these gains. For coding-agent harnesses on Terminal-Bench Lite, ceiling minus mismatch estimated with a stronger model as a stand-in for the optimum reproduces the accuracy ranking, while either term alone does not. Code is available at https://github.com/2020-qqtcg/HRP.

cs.LG · cs.CL

版本更新 · v2arXiv:2603.11137

Scaling Reasoning Efficiently via Relaxed On-Policy Distillation

Jongwoo Ko, Sara Abdali, Young Jin Kim, Tianyi Chen, Pashmina Cameron

原文摘要

On-policy distillation (OPD) is a promising approach for transferring reasoning capabilities to capacity-constrained models, yet sampled-token OPD often suffers from entropy collapse and negative transfer. We identify uniform credit allocation as a key bottleneck: existing objectives obtain dense teacher feedback on student-generated trajectories, but apply it uniformly across sampled tokens despite highly heterogeneous token-level signals. Most tokens carry redundant near-zero credit, while rare heavy-tailed negative credits can dominate updates and prematurely suppress plausible reasoning trajectories. We introduce REOPOLD (Relaxed On-Policy Distillation), a framework that relaxes this uniform assignment by controlling where teacher feedback is applied, how strongly it affects the update, and when the allocation rule changes during training. Across diverse reasoning tasks, REOPOLD improves sampled-token OPD over recent post-training baselines with up to 12x higher sample efficiency, and extends to cross-vocabulary distillation, self-distillation, and test-time scaling.

cs.LG · cs.CL

新论文 · v1arXiv:2610.37430

When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models

Ananth K Suresh, Aveen Dayal, Gouthaman KV, Aditya Pola, Vineeth N Balasubramanian

原文摘要

For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable questions is shaped by three sources of control: inference-time instructions, post-training, and internal activation interventions. Across three MLLMs on the MoHoBench benchmark, we systematically manipulate the meaning and placement of refusal instructions, compare supervised and preference-based post-training, and intervene on a refusal-related activation direction derived from naturally answered and refused examples. We find that refusal is highly sensitive to both the meaning and placement of instructions, with model-specific changes at the sample level that are obscured by aggregate refusal rates. Evidence-aware and general refusal instructions have stronger effects than an intentionally mismatched safety instruction. Post-training produces larger increases in refusal, but can also induce responses that combine refusal language with attempted answers. In contrast, activation intervention produces smaller yet transferable changes across prompt conditions, without evidence of indiscriminate refusal on tested visually answerable samples. These findings show that refusal is jointly shaped by external instructions, post-training, and internal representations, highlighting the need to disentangle evidence recognition from a model's learned propensity to refuse.

cs.LG · cs.CV

新论文 · v1arXiv:2610.34477

Finite-Sample Sensitivity of Kernel Metrics for Generative Evaluation

Tomasz Radzikowski, Mateusz Modrzejewski

原文摘要

Kernel distances such as the Kernel Inception Distance (KID) and the Kernel Audio Distance (KAD) compare generated and reference samples in an embedding space, and model comparisons rest on their values. Each ships with one fixed kernel, and a reported value does not say which changes that kernel detects at the sample size used. We run each default as a permutation two-sample test on real embeddings of symbolic music, audio and images, comparing a sample with an independent sample changed by a mean shift, an isotropic contraction, or a rotation that keeps the mean and the covariance trace and is matched to the shift in transport distance. At 780 samples per side, both defaults rarely detect the rotations in five evaluation spaces, while they detect the shifts in every run. The median-bandwidth Gaussian kernel of KAD detects the contraction and regains power against rotations on image embeddings with an order of magnitude more samples. The cubic kernel of KID contains second- and third-moment information but misses the contraction in most runs, and our experiments do not establish a high-power rotation regime for it over the tested sample sizes. Narrower Gaussian bandwidths restore sensitivity in music and audio, an $\ell_1$ Laplace kernel does so for images, and aggregated tests combine both at nominal level. We recommend that a reported kernel distance state its sample budget and the changes its configuration detects at that budget.

cs.LG · cs.SD

版本更新 · v2arXiv:2310.11324Google Research

Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting

Melanie Sclar, Yejin Choi, Yulia Tsvetkov, Alane Suhr

原文摘要

As large language models (LLMs) are adopted as a fundamental component of language technologies, it is crucial to accurately characterize their performance. Because choices in prompt design can strongly influence model behavior, this design process is critical in effectively using any modern pre-trained generative language model. In this work, we focus on LLM sensitivity to a quintessential class of meaning-preserving design choices: prompt formatting. We find that several widely used open-source LLMs are extremely sensitive to subtle changes in prompt formatting in few-shot settings, with performance differences of up to 76 accuracy points when evaluated using LLaMA-2-13B. Sensitivity remains even when increasing model size, the number of few-shot examples, or performing instruction tuning. Our analysis suggests that work evaluating LLMs with prompting-based methods would benefit from reporting a range of performance across plausible prompt formats, instead of the currently-standard practice of reporting performance on a single format. We also show that format performance only weakly correlates between models, which puts into question the methodological validity of comparing models with an arbitrarily chosen, fixed prompt format. To facilitate systematic analysis we propose FormatSpread, an algorithm that rapidly evaluates a sampled set of plausible prompt formats for a given task, and reports the interval of expected performance without accessing model weights. Furthermore, we present a suite of analyses that characterize the nature of this sensitivity, including exploring the influence of particular atomic perturbations and the internal representation of particular formats.

cs.CL · cs.AI · cs.LG