跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+1已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 03:03 · 全站最近入库：2026-11-08 06:02各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 04:03 — 2026-11-09 04:03（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：2026-11-09 03:03最近发现新增版本：2026-11-08 06:02库内论文最新更新时间：2026-11-06 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。

17,100 已收录论文14,661 AI 相关4,057 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

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 17,100 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 03:01（北京时间）
新论文 · v1arXiv:2610.39099

Data Driven Modeling of a Magnetic Newton's Cradle using the Optimization of a Discrete Loss

Dimitris Koumoutsakos, Sergey Litvinov

原文摘要

We investigate the dynamics of a magnetic Newton's cradle (MNC) using experiments, simulations through forward and inverse data driven models. In the MNC magnetic forces replace the instantaneous elastic collisions of the classical design. We compare point-dipole and finite-current-loop interaction models, identifying their parameters from direct measurements as well as through pendulum trajectories using the method of Optimizing a Discrete Loss (ODIL). For a synthetic three-pendulum twin, ODIL reproduces discretized trajectories to round-off accuracy and infers accurately magnet strength and damping from sparse data even when only one pendulum is observed. Model comparison shows that the point-dipole model compensates for near-field errors by underestimating the magnetic moment and produces structured residuals at each momentum transfer. The current-loop model removes this bias and, on full-precision recordings, also recovers the magnet radius. The continuous magnetic coupling reproduces the classical cradle's key behavior while enabling richer dynamics, including $N$ normal modes and a static repulsion offset. These results demonstrate that ODIL can infer physical parameters, and quantify model-form errors from sparse, noisy observations. Ultimately, this work establishes the magnetic Newton's cradle not merely as a physical curiosity, but as a rigorously quantifiable testbed for computational inverse modeling.

cs.CE · physics.class-ph · physics.comp-ph

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2606.14867

Evaluating the Robustness of Proof Autoformalization in Lean 4

Zhengtao Gui, Sheng Yang, Zhouxing Shi

原文摘要

Proof autoformalization aims to translate a mathematical informal proof in natural language into a formal proof in a formal language such as Lean 4. Existing LLM-based models for proof autoformalization are typically evaluated only on well-formed informal proofs from curated datasets. We argue that a robust proof autoformalizer must remain faithful even for informal proofs that diverge from these idealized ones, and we present the first study on the robustness of proof autoformalization models. We formulate two categories of perturbations: a global perturbation paraphrases the informal theorem-proof pair in a different style, under which the formalization should remain consistent; a local perturbation alters a value, symbol, or proof step, possibly in a counterfactual way, and a robust formalization should faithfully reflect the perturbation rather than reverting to the original or inferring a different one on its own. We build a benchmark with both perturbations on miniF2F and MATH-500, and automatically measure how stable a proof autoformalization's correctness is under global perturbations and how faithfully its output reflects local perturbations. We evaluate seven recent models: their correctness varies under global perturbations, and they mostly fail to remain faithful under local perturbations.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39093

Do Open-Source Video Ad Clients Follow VAST 4.3? An Empirical Study of Player Conformance

Prashant Chaudhary

原文摘要

Context. Video ads reach streaming and connected-TV apps through VAST documents, and companies count ads from the HTTP requests the player's client library sends. A tracking error can leave playback intact and surface later as disagreeing counts. Objective. To measure how far open-source VAST clients meet the player duties of VAST 4.3, whether verdicts depend on the runtime, what generated inputs add, and when deviations were introduced. Method. A catalog of 108 requirements was extracted from the specification. Six open-source clients in eight configurations ran against 69 crafted test vectors and the IAB's sample tags, under Node.js and in five browsers on three engines, with every request recorded. Grids, fuzzing and a sweep over 36 releases extend the catalog; Google's closed IMA SDK is a comparison. Results. The open-source clients have 78 findings and 3 to 21 must-level deviations per configuration. None of the tested open-source clients that follow redirect (Wrapper) chains fills the required macros in a Wrapper's request; the archived IMA 3.794.0 configuration tested does. One library's timeout throws an exception no caller can catch, ending a Node.js process. In a pod (ads played back to back), one player credits a finishing ad's completion to the next ad, fixed in release 5.3.0. A library can conform under jsdom and deviate in every tested Chrome- and Safari-family browser. Conclusions. The tested clients diverge on the duties other companies use to count and diagnose ads. A shared conformance suite run in real browsers could expose most graded deviations; the harness and vectors are released.

cs.SE · cs.MM · cs.NI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39091

Time-Switching Energy-Harvesting Quantum Relay-Assisted FSO Communication with Displaced Squeezed Signaling

Nizar Khalfet, Francisco A. Monteiro, Ioannis Krikidis

原文摘要

This paper investigates a quantum relay-assisted free-space optical (FSO) communication system in which a decode-and-forward relay is powered by time-switching optical energy harvesting and forwards the decoded information using displaced squeezed signaling (DSS). To obtain a physically con sistent description of turbulence-induced loss for non-classical signaling, each hop is modeled as a pure-loss bosonic chan nel whose random transmissivity follows a truncated Gamma Gamma law, i.e., the Gamma-Gamma irradiance conditioned on the physically admissible range of a transmittance. Under this channel, coherent states remain pure, whereas squeezed states are mapped to mixed Gaussian states. Under Helstrom-optimal detection, the average error of the coherent hop is derived in exact Laplace-series form. For the DSS hop, we derive the output states, a closed-form Gaussian fidelity with rigorous two sided bounds on the Helstrom error, and prove the existence of a loss-aware optimal photon partition. Furthermore, we prove that optimized squeezing preserves both the turbulence-induced diversity order and the leading asymptotic constant of coherent relaying, so that its benefit is a finite-photon-budget effect, and that an interior throughput-optimal harvesting fraction exists. A covariance-preserving benchmark with an exact series representation yields an analytical upper bound on the achievable throughput whose absolute gap vanishes as the relay photon budget grows. Numerical results based on the exact mixed-state Helstrom error quantify the benefit of squeezing and the regime in which it is useful.

quant-ph · cs.IT

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39090

PALATE: Pretrained Autoregressive LATent Embeddings

Sizhang Zhao, Praveen Gondhi, Bin Li, Utsaw Kumar

原文摘要

Local-commerce and food-delivery marketplaces run large ads businesses: across hundreds of thousands of stores, merchants pay to be surfaced on browse and search surfaces, and a ranker decides which ads each consumer sees. Ranking well demands strong per-consumer personalization, yet a consumer's raw interaction history is sparse and high-cardinality. Our production ranker had personalized only through hand-engineered signals: aggregate engagement counts, categorical embeddings, and order-agnostic pooled history, with no learned per-consumer sequence representation. We learn one by training a GPT-style generative model of the consumer's engagement sequence to predict their next few engagements, and read its last-token hidden state as the consumer embedding. That single frozen vector becomes one shared personalization input for every ad surface: a daily job refreshes it, and each ranker consumes it as an ordinary dense feature. Offline it is the largest single-feature gain in our ranking-AUC ladder, a relative Eval-AUC gain of +0.59% on a browse-driven feed and +0.57% on a query-driven search surface. Online, it shipped as one component of bundled ranker retrains on both surfaces: Discovery (revenue +0.79%, conversion rate +1.17%) and Search (revenue +0.94%, conversion rate +1.50%). Three design choices adapt the generative objective to a repurchase-heavy marketplace, and are the core of our contribution: an exploration-weighted loss that up-weights businesses new to the consumer over ones already engaged; a two-stage curriculum: a click pretrain that keeps the representation broad and exploratory, then a conversion fine-tune that sharpens purchase intent; and a designed negative set that pairs a fixed popular-business pool and businesses shown but not ordered with random and in-batch negatives.

cs.IR · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39088

Comparing wins above replacement across baseball history

Michael Doane

原文摘要

This paper introduces iWAR, a system for estimating wins above replacement from historical Retrosheet play-by-play records. We compare its runs-allowed (RA) and fielding-independent pitching (FIP) versions with Baseball-Reference WAR (bWAR), FanGraphs WAR (fWAR) and run-based alternatives. The common historical sample contains 1,471 team-seasons across 79 years and 1,441 consecutive franchise pairs. Under the fixed conversion to wins, same-season mean absolute error was 3.695 credited wins per 162 games for iWAR RA, compared with 4.724 for bWAR and 4.744 for fWAR: reductions of 21.793% and 22.112%. The paired differences were -1.030 and -1.049, with conditional 95% calendar-block intervals of [-1.367, -0.706] and [-1.398, -0.725]. Calibration preserved most of these advantages. A broader bWAR comparison across 118 years retained a smaller advantage; modern differences were smaller or sometimes reversed. PythagenPat reconstructed records more closely than every WAR total. Next-season replay gave nearly identical fixed-map errors for iWAR FIP and fWAR, while calibration changed their ordering. An amended chronological learned-model analysis of 1,552 franchise pairs gave neural RMSE 10.163, versus 9.946 for the development-selected ridge comparator (difference +0.217; conditional interval [0.015, 0.474]). This work provides a new, specialized framework for analyzing WAR across eras and provides context for future work using learning models.

cs.CE · stat.ME

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39086

HALO: Human-Grounded Granular Evaluation of Proactive Agents

Wenqi Pei, Xinyu Wang, Jinrui Liu, Cynthia Zhuge, Jie Sheng, Zihan Wang, Tenghao Huang, Muhao Chen

原文摘要

Proactive agents must decide whether, when, and how far to assist without an explicit request. Existing evaluations often rely on automated scores or binary verdicts, offering limited human diagnosis of evidence use and intervention quality. To address these gaps, we introduce HALO, Human-Grounded GrAnular EvaLuation of PrOactive Agents. HALO-Arena covers 4 proactive need types, Dependency, Blockage, Risk, and Opportunity, across 10 everyday domains and 50 personal environments. Scenarios require agents to retrieve evidence from structured databases or unstructured documents before judging whether assistance is warranted and choosing its scope. On HALO-Arena, we construct HALO-8K, a dataset of 8,000 responses from 10 models across 400 scenarios, annotated by 20 experts for action level, acceptability verdicts, and human diagnosis. For rejected responses, human diagnosis covers timing, intensity, and other issues. Recorded reasoning tokens and retrieval rounds measure the effort behind these responses. The best proprietary and open-source models achieve 72.88% and 56.63% acceptability, respectively. Our central finding is that over-proactivity is the dominant observed failure mode, with Premature timing or Excessive intensity accounting for 61.20% of rejected responses. Further analyses yield complementary findings on the balance between intervention and restraint, evidence acquisition, and the calibration of timing and intensity.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39084

Usable String Matching by Humans through Overlaid Presentation

Enka Blanchard, Ted Selker, Alan T. Sherman, Levi Gabasova, Mehdi Bouhamidi, Ryan Robucci

原文摘要

We explore the ability of humans to compare strings, an error-prone activity which occurs in critical contexts (e.g., bank wire transfers), where it is not always possible to trust a machine to help. We hypothesize that understanding human cognitive and perceptual limitations can improve how people verify whether two character strings are identical. We vary the strings' relative positions and presentation formats to measure comparison speed and accuracy. We present results from two online experiments with 178 human subjects who compared pairs of strings to determine whether the two strings were identical. The strings varied by length (5 characters to small paragraphs), difference (matching versus making a random change or a hard-to-detect one), alignment (vertical versus horizontal), and presentation (Overlay versus Adjacent). Our experiments test whether, by encoding and memory tasks into perceptual tasks, overlay formats can offer advantages in speed and accuracy. Results show that overlay offers a substantially faster method, with improved accuracy for long strings. Nevertheless, mean error rates were at least 4 percent, indicating that even the improved approach may be insufficiently accurate for many verification tasks.

cs.CR · cs.HC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2604.07424

An Analysis of Artificial Intelligence Adoption in NIH-Funded Research

Navapat Nananukul, Mayank Kejriwal

原文摘要

Understanding the landscape of artificial intelligence (AI) and machine learning (ML) adoption across the National Institutes of Health (NIH) portfolio is critical for research funding strategy, institutional planning, and health policy. The advent of large language models (LLMs) has fundamentally transformed research landscape analysis, enabling researchers to perform large-scale semantic extraction from thousands of unstructured research documents. In this paper, we illustrate a human-in-the-loop research methodology for LLMs to automatically classify and summarize research descriptions at scale. Using our methodology, we present a comprehensive analysis of 58,746 NIH-funded biomedical research projects from 2025. We show that: (1) AI constitutes 15.9% of the NIH portfolio with a 13.4% funding premium, concentrated in discovery, prediction, and data integration across disease domains; (2) a critical research-to-deployment gap exists, with 79% of AI projects remaining in research/development stages while only 14.7% engage in clinical deployment or implementation; and (3) health disparities research is severely underrepresented at just 5.7% of AI-funded work despite its importance to NIH's equity mission. These findings establish a framework for evidence-based policy interventions to align the NIH AI portfolio with health equity goals and strategic research priorities.

cs.AI · cs.CY · cs.MA

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2607.01033

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

Andrzej Szablewski, Gabriel Konar-Steenberg, Raffaello Fornasiere, Nikita Menon, Stefan Heimersheim

原文摘要

Model organisms (MOs) - language models trained to exhibit undesired or unnatural behaviours - are frequently used as testbeds for evaluating white-box interpretability techniques. Current MOs are typically constructed via post-hoc supervised fine-tuning (SFT). Prior research has shown that interpretability methods can easily identify hidden behaviours in these MOs. However, recent work suggests that such post-hoc training methods may make interpretability unrealistically easy. We investigate this claim by constructing a suite of 61 $\verb|OLMo2-1B|$- and $\verb|gemma-3-1b-it|$-based MOs trained with seven different techniques, including standard post-hoc SFT, post-hoc DPO, and more realistic integration of MO data into the OLMo post-training DPO phase. We also use these to train 96 further variants via knowledge distillation. We use these MO variants to benchmark activation oracles, activation steering, the logit lens, and the Jacobian lens. Our findings show that (i) MO interpretability depends strongly on training objective, target behaviour, model architecture, and training data generation pipeline; (ii) substantial variance remains even after controlling for differences in the strength of target behaviour expression; and (iii) both our more realistic $\textit{integrated training}$ and knowledge distillation often yield less interpretable MOs than standard post-hoc methods. Our results cast substantial doubt on the validity of current MOs as interpretability proxies and suggest actionable steps towards more robust benchmarks.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39074

HERA: Assessing an Autonomous Agent for Quantum Device Measurement

Zachary W. Gomez, Nathan Bittner, Jordan T. McCourt, Qingrui Cao, Sandesh S. Kalantre, Benjamin H. Alexander, Zhiren Zheng, Kenji Watanabe 等 10 位作者

原文摘要

Large language model agents have begun to carry out laboratory measurements, but their scientific conclusions are useful only if they hold up under scrutiny. Demonstrating this requires controlled studies because models are stochastic and the context a model is given steers its conclusions. Here we introduce HERA, an autonomous agent designed to carry out cryogenic transport measurements of quantum devices. HERA draws on a knowledge base curated by lab members that holds measurement protocols, software tools, and literature references. We assessed HERA with a blinded trial against an expert on the stacking order of two multilayer graphene devices, an ablation of its knowledge base, and hundreds of offline replays in which models designed the trial's central measurement without running it. HERA's conclusions matched the expert's sealed conclusions on both tested devices. The ablation and the replay studies measured the impact of the knowledge base across different models, showing that with it every model designed better measurements and reasoned more soundly. To test how readily HERA extends to a new problem, we tasked it with finding and characterizing superconductivity in a different material system. This required a few hours of expert time to develop a new experimental protocol, after which HERA did both autonomously. These results show that, with a knowledge base designed for the task, language model agents are capable of sound experimental science.

cond-mat.mes-hall · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39073

Certified Access: Test-Time Compute on Conservative Physics-Structured AI Memory

Pratik Jawahar, Riccardo Maggioni, Jonas Petersen, Matteo Cozzi, Gian-Alessandro Lombardi, Marek Gajewski, Camilla Mazzoleni, Federico Martelli

原文摘要

Test-time compute mechanisms, such as retries, confidence-based escalation and non-local routing, are conventionally deployed as black-box additions to trained models. They typically add forward passes or halting heads, but give no structural account of what the extra compute buys or what it costs against the model's stability guarantees. We therefore propose a theoretical and empirical framework for test-time compute on a Latent Dynamics Model used as the memory. When an encoder's latent state $(q,p)$ is advanced by a damped symplectic (velocity-Verlet) step of a learned Hamiltonian, every mechanism that expands access has an exact phase-volume Jacobian, a bounded energy change and an exact latch-transport law. We split reachability failures into escape failures and reach failures. A Lorentz squeeze resolves escape failures by injecting bounded, re-absorbable energy. A gated canonical translation (a one-step channel across $C_T$) resolves reach failures at $\detJ=1$ exactly. We verify this certificate stack on a designed analytic testbed. Reaching just beyond $C_T$ with the squeeze requires energy that grows quadratically in the excess distance. The gated channel's energy cost equals the potential difference between its endpoints, which is zero on the matched channel measured here, and so does not grow with distance in this testbed. An untrained state-replacing map annihilates phase volume and erases stored charge, while the gated channel transports it. On a learned physics-structured memory, we show calibrated compute rationing and evaluate the scaling limits of non-local routing against a physics-free baseline. The main contribution is the formalization and verification of the certificate stack, not benchmark performance.

cs.AI · hep-ph

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2603.22691

Rank-Aware CPU Reservation and In-Place Adjustment for MPI CFD on Kubernetes

Tianfang Xie

原文摘要

Non-uniform meshes in computational fluid dynamics (CFD) can leave some Message Passing Interface (MPI) ranks waiting while others determine execution time. On Kubernetes, reserving a full CPU for every rank can then restrict the admission of further simulations. We investigate rank-aware CPU reservation and in-place request adjustment using OpenFOAM on three Kubernetes testbeds. The framework lowers reservations without hard CPU caps and uses online MPI profiles to adjust each rank's CPU request while the solver runs. We reduced the per-job CPU reservation from 16.0 to 5.9 vCPU, with single-job wall-clock time increasing by at most 6%. With these reservations on twelve nodes, six co-located simulations reached 4.09 times the throughput of sequential execution. Matched controls showed no consistent benefit from proportional request shaping. Beyond the nominal CPU capacity, throughput collapsed with default busy-waiting; yielding waits softened the collapse. In one six-simulation run, the online controller performed 120 in-place resizes of running solvers with zero pod restarts; the pipeline reached 2.5 times the throughput of sequential execution. In a separate comparison on 16 physical cores, reallocating ranks from the far field to the near-wall zone reduced fixed-iteration wall-clock time by factors of 2.0 to 4.2 relative to the original concentric allocation, depending on mesh and iteration horizon. The reallocation also outperformed balanced Scotch partitions. The implementation requires no changes to OpenFOAM, the MPI library, or the Kubernetes scheduler.

cs.DC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39072

PhysicsNeMo-Mesh: GPU-Native Meshes and Simulation Cases for Physics AI

Peter D. Sharpe, Corey Adams, Negin Sobhani, Sanjay Choudhry

原文摘要

PhysicsNeMo-Mesh is the mesh module of the open-source PhysicsNeMo library. It provides mesh processing similar to PyVista, but on the GPU, in batches, and inside PyTorch's automatic differentiation. Like xarray for unstructured meshes, it attaches semantics to data: fields on points, cells, and whole meshes have names and shapes, and a simulation case keeps the complete PDE problem specification, including boundary conditions and their prescribed values. It addresses three problems that plague machine learning for physical simulation. First, GPU utilization is bottlenecked by I/O, because samples are very large relative to the compute per forward pass. Second, because traditional preprocessing is expensive, it is necessarily done once, during curation, rather than differently for each model and workflow at the point of use. Finally, advanced equivariant and mesh-native architectures require information that curation ordinarily discards. Fast GPU mesh processing makes it practical to preprocess at the point of use: on one GB200 GPU, mesh operations run $6.2\text{--}1{,}068\times$ faster than VTK with 72 threads, and loading a whole case is $8.0\text{--}88\times$ faster than reading its VTK files ($8.0\times$ when both read the same fields). Overlapped with training, this preprocessing costs $2.1\%$ of throughput for a $127\,\text{ms}$ Transolver step. So one dataset serves many models and workflows: in PhysicsNeMo's public recipe, one stored DrivAerML dataset feeds Transolver, GeoTransolver, and GLOBE. PhysicsNeMo-Mesh combines the speed of an intermediate array store with the flexibility of an archival format such as VTK. When shapes and wall conditions both vary, per-boundary conditions cut a heat-conduction error from $87\%$ to $6\text{--}9\%$. And by keeping field ranks and named boundaries, it retains the semantic structure that next-generation architectures will need.

physics.comp-ph · cs.CG · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.04262

Can I Take Another Dose? DoseBench: Evaluating LLM Decision-Making in OTC Dosing QA

Maroof Kousar, Jinsung Han, Mohit Chandra, Munmun De Choudhury, Yibo Hu

原文摘要

Large language models (LLMs) are increasingly used to answer everyday health questions online, including whether someone can take another dose of an over-the-counter (OTC) medication. Yet this common safety-relevant setting remains underexplored in medical QA. Answering these questions requires checking previous doses, timing, and product-specific dosing limits, while recognizing when important information is missing. We introduce DoseBench, a focused benchmark of 230 consumer-style acetaminophen and ibuprofen dosing questions with human-annotated YES, NO, and AMBIGUOUS labels, and evaluate eight LLMs with five responses per question. Overall accuracy ranges from 44.8% to 62.2%, while models incorrectly approve 14.0% to 60.5% of prohibited doses. Models with similar accuracy can therefore make very different dosing decisions. Additional analyses show that these failures are not easily fixed by giving models more guidance. Asking models to check dose, timing, and missing information makes them more cautious, while providing drug facts improves accuracy but can make them overlook missing information. These findings highlight a broader challenge in health QA: applying rules correctly and recognizing when information is missing.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39070

When the Same Prompt Gives Different Answers (and the Same Answer Isn't Always Right): A Taxonomy of Failure Mechanisms, Testing, and Mitigation Strategies for LLM-Assisted Evidence Synthesis Read-and-Evaluate Tasks

Carole Lunny, Ian Goodman, Stefano Berrone, Connor TA Brenna, Emma K Reid, Ivan Perez-Neri, Shailesh Kolekar, Nityanand Jain 等 15 位作者

原文摘要

LLM read-and-evaluate tasks include study screening, data extraction, risk-of-bias assessment, and critical appraisal in the evidence synthesis field. These task differ in the amount of LLM interpretation they require. Some tasks involve retrieving relatively simple information, such as an author name, sample size, or publication year. Others require the LLM to read source material, interpret it against predefined criteria, and make a judgement. Researchers should assess whether LLM judgements are accurate, supported by appropriate source evidence, and reproducible under specified conditions. A correct judgement does not necessarily indicate that the underlying evaluation process is reliable. We developed a taxonomy for identifying mechanisms that may compromise these properties and for informing the selection and reporting of appropriate evaluation methods. The taxonomy outlines potential ways that LLM judgments from read-and-evaluate tasks can fail, called TRAILS (Taxonomy of Reliability And Instability in LLM-assisted evidence Synthesis). The TRAILS taxonomy addresses two complementary questions. First, what mechanisms can cause the same read-and-evaluate task to produce different judgements across repeated LLM runs? Second, what mechanisms can systematically affect the judgement or required output in a way that is not apparent from the judgement itself? This study presents 38 mechanisms and identifies the deployment settings in which each can operate, approaches for testing them, and mitigation strategies. We also show why repeatability or accuracy does not rule out a hidden judgement failure and why human evaluation should assess whether a judgement is grounded in the supplied evidence, independent of inappropriate external influence, and produced through a process that can be examined and verified.

cs.AI · cs.PF · stat.ME

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39068

Closed Gates Leak: Decomposing Adjacent-Layer Cancellation in Gated Language Models

Harvey Dam, Ganesh Gopalakrishnan

原文摘要

In transformer language models with gated multilayer perceptrons (MLPs), adjacent layers write partly against each other into the residual stream. Previous work interpreted this cancellation as layers proposing features and the next layer selectively rejecting the inappropriate ones. We decompose the cancellation by removing the previous layer's write from the input of the next MLP: what the next layer writes along that write splits into what it writes anyway and how much less its MLP writes because the write is present, which we call the damping, and we trace the damping to the neurons that carry it. In gated MLPs the damping is carried by neurons whose gates are closed. SiLU and GELU are slightly negative below zero, so a closed gate still passes a small negative multiple of what the neuron reads, and a neuron whose output weight points along its input weight then subtracts a fraction of the direction it read. Summed over such neurons, closed gates contract the average direction a gated MLP reads, and the previous layer's write is not singled out: the same layer's write at another token is damped as much or more, and the eigenvalue signature from which that work concluded that the correction is selective appears for another token's write as well. Over training the damping grows while what the next layer writes anyway stays flat, and cancellation begins when the damping overtakes it; annealing the negative side of the gate activation away in continued pretraining removes the damping and the cancellation. Adjacent-layer cancellation is therefore weak evidence about which features a model rejects: it is produced by a contraction that gated MLPs apply, on average over directions, to whatever they read.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2604.03529

Incentives Shape How Humans Co-Create with Generative AI

Nathanael Jo, Manish Raghavan

原文摘要

Generative AI is quickly becoming an integral part of people's everyday workflows. Early evidence has shown that while generative AI can increase individual-level productivity, it does so at the cost of collective diversity, potentially narrowing the set of ideas and perspectives produced. Our research stands in contrast to this concern: through two pre-registered randomized control trials, we show that incentives can moderate AI's homogenizing force in two creative writing tasks where participants can use AI interactively. Participants rewarded for originality relative to peers produce collectively more diverse writing than those rewarded for quality alone. These participants continued to engage with AI, but ultimately incorporated less AI-generated text verbatim. Our results reveal that the effects of generative AI depend not only on the technology itself, but also the behavioral strategies and incentive structures surrounding its use. We release a rich, anonymized dataset capturing fine-grained trajectories of human-AI co-creation across both studies.

cs.HC · cs.AI · cs.CY

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2608.10319

Do Personalized Skills Help Coding Agents? An Empirical Study of Developer Interaction Traces

Shuyan Huang, Kai Du, Andrew Lan

原文摘要

As human developers increasingly collaborate with large language model (LLM)-powered coding agents, adapting to individual developer preferences, potentially via agent skills, may offer way to reduce repeated guidance and improve collaboration efficiency. However, existing approaches primarily focus on building skills for each task rather for each human developer. In this work, we propose a framework that combines rule-based bootstrapping with LLM-based refinement to distill developer-specific skills from their interactions with coding agents. We further develop an interactive replay framework with an LLM-based developer simulator for evaluation. We conduct experiments on a collection of replayable real-world developer-agent interaction sessions. We find that overall, task-specific skills are the most effective; However, personalized skills become more effective when developer preferences are manifested with enough frequency, especially when developers work on tasks that are highly related. These findings provide empirical insights suggest that i) broadly transferable procedural knowledge can be more robust than developer-specific preference signals and ii) effective personalization requires mining from rich user preference signals.

cs.SE · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2602.00414

Toward Autonomous Laboratory Safety Monitoring with Vision Language Models: Learning to See Hazards Through Scene Structure

Trishna Chakraborty, Udita Ghosh, Aldair Ernesto Gongora, Ruben Glatt, Yue Dong, Jiachen Li, Amit K. Roy-Chowdhury, Chengyu Song

原文摘要

Laboratories are prone to severe injuries from unsafe actions, yet continuous manual safety monitoring, beyond mandatory pre-lab safety training, is infeasible. Vision-language models (VLMs) offer promise for autonomous laboratory safety monitoring, but their effectiveness in realistic settings is unclear due to the lack of visual evaluation data. Safety incidents are mostly documented as text, and existing studies are largely limited to textual understanding. Since hazardous scenes are unsafe to stage or capture, synthetic visual data becomes the only practical way to study this problem. This motivates us to introduce a structured data generation pipeline that converts textual laboratory scenarios into aligned quadruples using LLM as scene graph architect and image generation model as renderer, yielding \dataset: a synthetic dataset of 1,207 $\langle$image, scene graph, ground truth, hazard relevance mask$\rangle$ quadruples across 362 unique scenarios. Experiments on \dataset and seven open- and closed-source models show that VLMs perform effectively given a textual scene graph but degrade substantially in visual-only settings (i.e., where no scene graph is involved), indicating difficulty in extracting structured object relationships directly from pixels. We thus propose \saga, a post-training framework that leverages reinforcement learning to first structure the visual scene as a scene graph, and then predict hazards while aligning attention with relevant image regions. Evaluations with Qwen 3.5 and Gemma 4 show that our method nearly doubles the $F_1$ score over the baseline on our synthetic dataset and achieves an improvement of $\sim 3\times$ on a mini real-world dataset, while preserving general visual understanding in visual question answering.

cs.CV · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39066

DeepDiff: Neuro-Dynamic Regression-Freedom Checking for Security Patch Validation

Laboni Sarker, Shuvendu K Lahiri

原文摘要

Automatically validating AI-generated security patches remains a significant challenge. Eliminating a reported crash and passing existing test suites do not provide sufficient confidence that a patch correctly addresses the underlying vulnerability without introducing regressions. We present DeepDiff, a technique for regression-freedom checking that differentially evaluates candidate patches against the original vulnerable program using non-crashing inputs that exercise the affected code paths. The expectation of regression-freedom is that a correct patch should eliminate the vulnerability triggering crash while preserving the intended behavior on non-crashing inputs. To support this evaluation, we introduce a dataset of correct and incorrect patch candidates for open-source memory-safety vulnerabilities from CyberGym. We explore different approaches for non-triggering test input generation to reach the patched code, a purely neural regression-freedom checker, and a hybrid approach that combines dynamic analysis with neural techniques. Our experimental results demonstrate the effectiveness of regression-freedom checking in validating security patches while outlining the challenges that remain.

cs.CR · cs.SE

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39065

Tabular foundation models are not a shortcut to calibrated clinical risk prediction: a pre-registered simulation against the true risk

Ebrahim Khaled Ebrahim

原文摘要

Tabular foundation models are pretrained networks that predict risk from a table without fitting a model. Whether their risks are correct (calibrated) has been judged on real data, where true risk is unknown. In a pre-registered simulation with known true risks, we compared TabPFN-3.5, TabICLv2 and Mitra-v2 with Firth logistic regression with intercept correction (FLIC) in 58 scenarios and 1,000 random designs, then on four clinical data sets. Foundation models were not a shortcut to calibration. Their calibration slope matched FLIC's within the pre-registered margin in 36 or 37 of 48 core scenarios and was better only at n = 200 with a 3% event rate. Even 3,000 patients were often not enough: with FLIC, TabPFN-3.5 or Mitra-v2, 63%-83% of samples gave risks close to the truth in average, intercept and slope at event rates of 10% and 30%, and 35%-50% at 3%. TabICLv2 did worse (8%-39% and 1%-3%). Across the random designs, foundation models had a lower true calibration error than FLIC beyond the margin in 25%-38% and a higher one in none, though small losses were predicted at n = 3000. On real data, no model resisted a change in event rate between populations; recalibration on 500 local patients largely repaired it.

stat.ME · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39062

HEARSAYBENCH: Can AI Navigate from Abstract Human Rights to Lived Lives?

Ava Iranmanesh, Sobhan Lotfi, Ali Iranmanesh, Liwei Jiang

原文摘要

As large language models (LLMs) reach users worldwide, personal counseling has emerged as one of their most consequential applications, sometimes in life-critical situations. Yet broader access does not mean equal support. The text these models learn from disproportionately reflects people with the freedom to act, embedding an implicit assumption that users can leave, report, seek help, or appeal. Existing benchmarks largely inherit this assumption, overlooking users whose agency is constrained by structural conditions such as war, statelessness, and displacement. Meanwhile, models are increasingly shaped by normative frameworks such as Constitutional AI, often grounded in ideals like human rights, even as many users live where those rights offer little protection or recourse. We introduce HEARSAYBENCH, a human-authored dataset of 400 test cases with in-region human verification grounded in scenarios where structural barriers stand between a person and the exercise of their rights. Built from United Nations and other human rights reports, the scenarios span over 80 regions and cover social, environmental, and personal constraints. Drawing on Nobel laureate Amartya Sen's Capabilities Approach, HEARSAYBENCH tests whether models can distinguish what a person is formally entitled to do from what they are actually able to do. Across 14 proprietary and open-weight models, every model infers a user's situation more accurately than it advises on it. Stronger models narrow this gap, but their recommendations remain unsafe to follow. Models may understand lives shaped by constraint, yet that knowledge remains hearsay if it cannot translate into advice people can act on. Those people are not alignment's long tail; they are among the users for whom good guidance matters most. HEARSAYBENCH makes this failure measurable, a step toward systems that serve the world as it is rather than as assumed.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39060

Tracing the Influence of Retrieved Literature on Scientific Ideation in LLMs

Zhipin Wang, Koel Dutta Chowdhury, Christoph Leiter, Steffen Eger

原文摘要

Large language models (LLMs) are increasingly used to generate research ideas from scientific literature. Existing systems combine literature retrieval with search, generation, and refinement, making it difficult to determine how the retrieval strategy itself shapes the ideas generated. We isolate this contribution by comparing five retrieval strategies and a No Literature baseline under a shared generation protocol, using 45,000 ideas across 15 machine learning topics and ten open-weight models. Relative to no literature basline, all five retrieval strategies reduce diversity among ideas generated from the same set of papers by $5.6\%$--$17.9\%$, while most increase diversity by $10.2\%$--$18.1\%$ when ideas from different sets are pooled. Effects on idea quality and novelty are modest on average but vary across topics. We further apply a JSD-based attribution method to trace the influence of individual input papers on each idea's motivation, core idea, approach, and validation. In $79.3\%$ of ideas, only one or two distinct papers receive the highest attribution scores across these four parts. These leading sources often recur across ideas generated from the same set of papers, and groups with more varied source attribution contain more diverse ideas ($r=0.70$). These findings motivate various sets of literature and the use of attribution to guide the choice of sources for subsequent ideas.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2505.00311

A Practical GPU-Enhanced Matrix-Free Primal-Dual Method for Large-Scale Conic Programs

Zhenwei Lin, Zikai Xiong, Dongdong Ge, Yinyu Ye

原文摘要

In this paper, we introduce a practical GPU-enhanced matrix-free first-order method for solving large-scale conic programming problems, which we refer to as PDCS, standing for the Primal-Dual Conic Programming Solver. Problems that it solves include linear programs, second-order cone programs, convex quadratic programs, and exponential cone programs. The method avoids matrix factorizations and uses sparse matrix-vector multiplication as its core operation, which is memory-efficient and well suited for GPU acceleration. It builds on a variant of the primal-dual hybrid gradient method (PDHG) called reflected restarted Halpern PDHG with averaging, which we prove converges on general conic programs, and further incorporates several practical enhancements. Its GPU implementation, cuPDCS, uses customized computational schemes that assign cones of different types and sizes to different levels of the GPU architecture. Numerical experiments demonstrate that cuPDCS is generally more efficient than state-of-the-art commercial solvers and other first-order methods on large-scale conic program applications, including Fisher market equilibrium problems, Lasso regression, and multi-period portfolio optimization. cuPDCS also exhibits better scalability, efficiency, and robustness compared to other first-order methods on the conic program benchmark dataset CBLIB. These advantages are more pronounced in large-scale settings.

math.OC · cs.MS

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39056

When Does Closed-Loop Curriculum Adaptation Help GRPO? An Analysis on Countdown

Yuv Raj Pant, Moshiur Rahman Tonmoy, Kaiqun Fu

原文摘要

Reinforcement fine-tuning has become an effective post-training paradigm for improving the reasoning ability of large language models, but it remains sample-inefficient when the training distribution is mismatched to the model's current capability. In group-relative objectives such as Group Relative Policy Optimization (GRPO), problems that are too hard often produce uniformly failed rollouts, while problems that are too easy provide little useful advantage signal. In this work, we introduce CLOSEDLOOPGAUSSIAN, a lightweight adaptive curriculum method for efficient reinforcement fine-tuning on ordered reasoning tasks. Instead of following a fixed easy-to-hard schedule, CLOSEDLOOPGAUSSIAN preserves the structure of an Easy-to-Hard Gaussian sampler while dynamically adjusting its pace and spread using periodic probe estimates of per-difficulty accuracy. The controller accelerates training when the model is ahead of a reference curriculum, slows down when the model falls behind, and broadens sampling when earlier difficulties regress. This feedback mechanism keeps training focused on tasks that are challenging but still learnable, without changing the reward function, model architecture, or underlying GRPO optimizer. Experiments on symbolic arithmetic reasoning task show that CLOSEDLOOPGAUSSIAN improves exact-match accuracy over vanilla GRPO and fixed open-loop curricula, with the strongest gains in intermediate-difficulty regimes where group-relative learning signals are most informative. Codes are available at https://github.com/yuvrajpant56/Closed_Loop_Curriculum.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.25752

Which Answers Agree? Evaluating Self-Consistency from Incomplete Logs

Zhengtao Gui, Zexi Chen, Weining Shen

原文摘要

Self-consistency answers a question by generating multiple responses and selecting the most frequent answer. Its evaluation requires both correctness labels and information about which responses agree; correctness labels alone generally determine neither voting accuracy nor the benefit of additional votes. For tasks in which all correct responses map to the same normalized answer, we derive sharp accuracy bounds using classical stochastic majorization, refined by abstention counts and answer-category limits. We also derive sharp bounds on accuracy changes across voting budgets that share a response pool. When answer-category counts are available, we introduce Plurality-U, a count-based dynamic program that exactly averages voting accuracy over all fixed-size subsets, eliminating finite-pool Monte Carlo error. On 1,403,520 stored outputs from recent reasoning models in Scorio Lite and Scorio Trace, recording abstention counts and answer options cuts the width of SuperGPQA accuracy intervals by more than half. For moving from 9 to 17 votes, our bounds reduce interval width by a median of 77.7% on SuperGPQA and 72.3% on 14 GPQA and Adaptive-Consistency pools, compared with a baseline that combines separate accuracy bounds. Exact computation has lower paired median runtime than 2,000-subset Monte Carlo in 14 of 15 tested pool-budget settings. These results clarify what incomplete logs can establish and provide an exact evaluation procedure when answer counts are retained.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39055

Controlling Continuum Robot Shape in Enclosed Space with Embedded Sensing using IMU and Actuator Feedback

Jiahe Wang, Sultan Haidar Ali, Eric Weissman, Jiefeng Sun

原文摘要

Controlling continuum robots' shape to operate in confined, enclosed environments is challenging, since it cannot rely on external observation. Existing embedded shape sensing methods, although they can provide shape reconstruction, are insufficient to enable real-time shape control due to limitations in reconstruction accuracy, computational speed, and compatibility with shape-state representation. In this work, we present an embedded shape-sensing framework that uses readily available IMU and actuator feedback to reconstruct the continuum-robot shape with a neural network, providing accurate shape feedback at $400~\mathrm{Hz}$ for physical closed-loop control. We systematically investigate four key reconstruction design choices: shape-state representation, network architecture, training objective, and sensing inputs, through more than $5{,}000$ trained models on one- and three-segment continuum robots. The selected framework achieves global shape errors of $0.9\%L$ and $1.7\%L$ (of the robot length) for the one- and three-segment systems, respectively. In closed-loop control, one-segment tip tracking achieves a $6.1\%L$ error and three-segment shape control achieves a $2.3\%L$ global shape error, whereas IMU--PCC reconstruction using the same uncompensated measurements fails. Enclosed-space experiments further demonstrate operation without external observation. These results provide a practical framework for translating readily available embedded sensing into high-rate feedback for continuum robot closed-loop control.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39054

Robotic Data Inoculation

Qiyang Li, Zhiyuan Zhou, Charles Xu, Sergey Levine

原文摘要

Modern robot learning systems feed on high-quality demonstration data, and their performance hinges on coverage. Without good coverage, small errors compound and the policy can quickly go into out-of-distribution states from which it cannot recover. Yet there is no principled way of collecting high-coverage data, and data collection often relies on heuristic instructions to human operators. Purely online reinforcement learning (RL) explores at the cost of safety and sample efficiency, while DAgger-style interventions require the policy to fail by chance before a human can correct it. In this paper, we ask instead whether we can modify how the robot behaves during data collection to collect more useful data. We propose Data Inoculation (DI), a shared-control framework in which a human provide closed-loop teleoperation commands while a residual policy perturbs these commands. The residual policy is trained online, on the growing dataset, with the same reward-maximizing objective later used for policy extraction, so its perturbations steer data collection toward the deviations a policy trained on the data would make. These perturbations expose the robot to a wider range of states, while the human's continuous corrections provide recovery demonstrations, "inoculating" the dataset against the learner's future mistakes while the human remains in control. On three challenging real-world robotic tasks, DI consistently improves or matches the success rate of policies trained with both behavioral cloning and offline RL, showing DI as an effective data collection strategy for improving downstream policy performance.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39050

BASILISC: Bayesian Alignment of Siamese-Integrated Latents for Inferring Single-cell Correspondence

Sepideh Saran, Mahsa Ghanbari, Uwe Ohler

原文摘要

Single-cell assays study the genome one cell at a time. Unimodal assays measure one molecular layer, the newer multimodal assays measure several. Most data are single-modality, so integrating them relies on models trained on multimodal assays, where the cell correspondence across modalities is known. These models return a matching, but rarely a confidence measure. We present BASILISC (Bayesian Alignment of Siamese-Integrated Latents for Inferring Single-cell Correspondence), a framework for matching single-cell representations with uncertainty on the correspondence itself. A BASILISC model is a probabilistic siamese network over the two modalities with a predictive distribution over the correspondence for a cross-modal pair of cells. We instantiate it at two levels: at the cell-type level the question is whether two cells share a cell type, at the cell level which cell is the partner in the other modality. To our knowledge, this is the first uncertainty score on a learned cross-modal correspondence evaluated against a known ground-truth pairing. BASILISC learns a similarity across modalities that reflects cell lineage: an unseen cell type is matched to its lineage relative, matching still works on such held-out types (balanced accuracy 0.86 on unseen against 0.93 on seen types), and the predictive uncertainty rises on these out-of-distribution cells. The uncertainty score also selects high-quality matches for downstream analysis: discarding the least certain fifth of the pairs raises accuracy from 0.93 to 0.98. At the cell level, a deep ensemble of the deterministic matcher reaches 0.062 matching accuracy on the benchmark test set, above the competition winner; on the Bayesian matcher's assignment, the posterior flags the wrong matches that cross a cell-type boundary better than a deterministic read of the same scores (AUROC 0.68 against 0.64).

cs.LG

查看论文
arXiv ↗PDF ↗