120 2026-10-27T20:01:06+00:00 rss-help@arxiv.org python-feedgen cs updates on the arXiv.org e-print archive. oai:arXiv.org:2610.35869v1 2026-10-27T20:00:00+00:00 Albert Heekwan Kim arXiv:2610.35869v1 Announce Type: new Abstract: AI applications often invoke a model before they determine whether inference is necessary. That default is useful for ambiguous generative tasks, but it is poorly matched to work that is structured, deterministic, policy-bound, lookup-oriented, or directly verifiable. KORA introduces deterministic-first execution control: a pre-inference control layer that represents requests as task paths, attempts deterministic or structured routes first, validates route outcomes, records telemetry, and escalates to a model-candidate path only when deterministic handling is insufficient. This paper presents KORA as a deterministic-first execution-control architecture and reports bounded evidence from reproducible benchmark and validation workloads. In a reproducible 100-task deterministic-heavy benchmark, KORA-controlled execution avoided 80 of 100 simulated model invocations versus a naive direct baseline, with zero mismatches in the deterministic subset. The workload is intentionally deterministic-heavy: it includes categories such as FAQ lookup, policy/rule checking, schema or format validation, deterministic transformation, and a smaller set of model-needed ambiguous or generative tasks. A separate synthetic customer-support local/no-network validation workload demonstrates route measurement and avoided model-call accounting in an application-shaped setting. These results support bounded model-invocation reduction and validation visibility in documented workloads. They do not establish production behavior, real-provider outcomes, energy outcomes, API-cost reduction, or broad superiority.

arXiv:2610.35869v1 Announce Type: new Abstract: AI applications often invoke a model before they determine whether inference is necessary. That default is useful for ambiguous generative tasks, but it is poorly matched to work that is structured, deterministic, policy-bound, lookup-oriented, or directly verifiable. KORA introduces deterministic-first execution control: a pre-inference control layer that represents requests as task paths, attempts deterministic or structured routes first, validates route outcomes, records telemetry, and escalates to a model-candidate path only when deterministic handling is insufficient. This paper presents KORA as a deterministic-first execution-control architecture and reports bounded evidence from reproducible benchmark and validation workloads. In a reproducible 100-task deterministic-heavy benchmark, KORA-controlled execution avoided 80 of 100 simulated model invocations versus a naive direct baseline, with zero mismatches in the deterministic subset. The workload is intentionally deterministic-heavy: it includes categories such as FAQ lookup, policy/rule checking, schema or format validation, deterministic transformation, and a smaller set of model-needed ambiguous or generative tasks. A separate synthetic customer-support local/no-network validation workload demonstrates route measurement and avoided model-call accounting in an application-shaped setting. These results support bounded model-invocation reduction and validation visibility in documented workloads. They do not establish production behavior, real-provider outcomes, energy outcomes, API-cost reduction, or broad superiority.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35870v1 2026-10-27T20:00:00+00:00 Shigemi Ishida, Ryo Takahashi, Satoshi Ido arXiv:2610.35870v1 Announce Type: new Abstract: Near-field communication (NFC) sensor tags are promising for low-cost, scalable environmental monitoring due to their battery-free, maintenance-free operation. Recent advances in NFC reader garments, integrated with conductive textile coils, allow users to interact seamlessly with tags scattered across their daily environment. However, dynamically identifying and reading out a fluctuating number of tags remains challenging. If multiple tags speak and collide, they must randomly retry until an empty slot is found. Standard NFC identifies tags using time-slotted ALOHA, which assigns tags to fixed time slots. If multiple tags speak and collide, they must retry with a tag filter to avoid collision until all tags are identified. This rigid setup is fragile in dynamic settings; as tags enter and leave the communication range, the heavy overhead of slot re-negotiation and inevitable collisions restricts data throughput. Here, we propose an asynchronous NFC based on packet interval modulation. By treating minimal ID-only packets as pulses and encoding data in their intervals, we enable communication through silence. This ensures the channel remains mostly unoccupied, drastically reducing collisions even in high-density environments. Our simulations show that for ten tags, our method achieves a 4-fold increase in throughput compared to the ALOHA approach. Furthermore, the system remains robust, maintaining over 160 bps/tag throughput even under extreme fluctuations in the number of tags over 20.

arXiv:2610.35870v1 Announce Type: new Abstract: Near-field communication (NFC) sensor tags are promising for low-cost, scalable environmental monitoring due to their battery-free, maintenance-free operation. Recent advances in NFC reader garments, integrated with conductive textile coils, allow users to interact seamlessly with tags scattered across their daily environment. However, dynamically identifying and reading out a fluctuating number of tags remains challenging. If multiple tags speak and collide, they must randomly retry until an empty slot is found. Standard NFC identifies tags using time-slotted ALOHA, which assigns tags to fixed time slots. If multiple tags speak and collide, they must retry with a tag filter to avoid collision until all tags are identified. This rigid setup is fragile in dynamic settings; as tags enter and leave the communication range, the heavy overhead of slot re-negotiation and inevitable collisions restricts data throughput. Here, we propose an asynchronous NFC based on packet interval modulation. By treating minimal ID-only packets as pulses and encoding data in their intervals, we enable communication through silence. This ensures the channel remains mostly unoccupied, drastically reducing collisions even in high-density environments. Our simulations show that for ten tags, our method achieves a 4-fold increase in throughput compared to the ALOHA approach. Furthermore, the system remains robust, maintaining over 160 bps/tag throughput even under extreme fluctuations in the number of tags over 20.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35871v1 2026-10-27T20:00:00+00:00 Mayank Atreya, Navin Chhibber, Shilpa Ajay, R Loshini, K Anitha arXiv:2610.35871v1 Announce Type: new Abstract: Smart retail systems now depend on the application of real-time video analytics, which can track customer behavior, manage shelf inventory, and enhance in-store processes. Currently cloud-based applications of video analytics face multiple challenges such as high latency of the communication itself, privacy concerns surrounding transmitting raw video, and the computational demands of the heavy vision models used on the often limited network infrastructure. These challenges lead to delays in decision making where prompt action is crucial. To address these short comings, this research proposes Fog-based analytics withRetailLightNet (F RLN), an ultra lightweight deep learning model which is specifically suited for the edge, resource constrained hardware capabilities of devices like the NVIDIA Jetson Nano. The approach utilizes publicly available retail datasets, SKU110K, and RPC, which contains an assortment of product, shelf, and customer interaction scenarios for training.The processes include preprocessing images through normalization and employing motion-gated frame selection to improve efficiency by omitting redundant frames to ease the workload on the fog processing. Internal within Retail LightNet, image feature extraction is realized through depth wise separable convolutions which help support spatial and channel-wise learning without heavy computational costs. The model employs two prediction heads presence classification, and density regression, supporting shelf monitoring and queue assessment. The proposed F RLN framework attained better results in terms of precision (0.988), recall (0.976), and F1-score(0.976) when compared with existing convolutional neural network. Keywords: fog computing, motion-gated frame selection,NVIDIA jetson, retail lightnet, smart retail.

arXiv:2610.35871v1 Announce Type: new Abstract: Smart retail systems now depend on the application of real-time video analytics, which can track customer behavior, manage shelf inventory, and enhance in-store processes. Currently cloud-based applications of video analytics face multiple challenges such as high latency of the communication itself, privacy concerns surrounding transmitting raw video, and the computational demands of the heavy vision models used on the often limited network infrastructure. These challenges lead to delays in decision making where prompt action is crucial. To address these short comings, this research proposes Fog-based analytics withRetailLightNet (F RLN), an ultra lightweight deep learning model which is specifically suited for the edge, resource constrained hardware capabilities of devices like the NVIDIA Jetson Nano. The approach utilizes publicly available retail datasets, SKU110K, and RPC, which contains an assortment of product, shelf, and customer interaction scenarios for training.The processes include preprocessing images through normalization and employing motion-gated frame selection to improve efficiency by omitting redundant frames to ease the workload on the fog processing. Internal within Retail LightNet, image feature extraction is realized through depth wise separable convolutions which help support spatial and channel-wise learning without heavy computational costs. The model employs two prediction heads presence classification, and density regression, supporting shelf monitoring and queue assessment. The proposed F RLN framework attained better results in terms of precision (0.988), recall (0.976), and F1-score(0.976) when compared with existing convolutional neural network. Keywords: fog computing, motion-gated frame selection,NVIDIA jetson, retail lightnet, smart retail.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35872v1 2026-10-27T20:00:00+00:00 Harikrishna Kuttivelil, Lakshmi Krishnaswamy, Srikar Rani, Psi Padhya, Alex Woelkers, Vasavi Doddamani, Nayan Bhatia, Katia Obraczka arXiv:2610.35872v1 Announce Type: new Abstract: As connected applications -- from intelligence-enabled assistants and the smart homes to autonomous vehicles and large-scale Internet-of-Things deployments -- become increasingly pervasive, their performance, safety, and reliability depend more than ever on the underlying network. Designing and validating such systems therefore requires development environments that holistically account for both application and network behavior. While network simulators and emulators provide high-fidelity models of network behavior, connecting full-fledged applications to them remains difficult due to platform constraints, simulator-specific interfaces, and narrow single-purpose integrations. To address this gap, we present NSB2 (Network Simulation Bridge 2), an open-source network co-simulation middleware that relays payloads between application processes through a network simulator. NSB2 is composed of a central daemon and thin, embedded application- and simulator-side libraries for flexible integration, disentangled communication channels for unencumbered inter-process communications, and in-memory key-value storage for efficient payload caching and retrieval. NSB2 is designed to be lightweight, accessible, and extensible: it exposes streamlined interfaces supported in Python and C++, modular internal abstractions that let users swap or extend network simulation/emulation backends, flexible and configurable deployment modes, and integration helpers for the widely-used ns-3 and OMNeT++ network simulators. Our evaluation across multiple co-simulation scenarios shows that NSB2 introduces negligible memory and latency overhead, enabling researchers and developers to achieve richer, more realistic modeling of connected systems with minimum overhead.

arXiv:2610.35872v1 Announce Type: new Abstract: As connected applications -- from intelligence-enabled assistants and the smart homes to autonomous vehicles and large-scale Internet-of-Things deployments -- become increasingly pervasive, their performance, safety, and reliability depend more than ever on the underlying network. Designing and validating such systems therefore requires development environments that holistically account for both application and network behavior. While network simulators and emulators provide high-fidelity models of network behavior, connecting full-fledged applications to them remains difficult due to platform constraints, simulator-specific interfaces, and narrow single-purpose integrations. To address this gap, we present NSB2 (Network Simulation Bridge 2), an open-source network co-simulation middleware that relays payloads between application processes through a network simulator. NSB2 is composed of a central daemon and thin, embedded application- and simulator-side libraries for flexible integration, disentangled communication channels for unencumbered inter-process communications, and in-memory key-value storage for efficient payload caching and retrieval. NSB2 is designed to be lightweight, accessible, and extensible: it exposes streamlined interfaces supported in Python and C++, modular internal abstractions that let users swap or extend network simulation/emulation backends, flexible and configurable deployment modes, and integration helpers for the widely-used ns-3 and OMNeT++ network simulators. Our evaluation across multiple co-simulation scenarios shows that NSB2 introduces negligible memory and latency overhead, enabling researchers and developers to achieve richer, more realistic modeling of connected systems with minimum overhead.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35873v1 2026-10-27T20:00:00+00:00 Muhammad Dikko Gambo, Sami Ul Haq, Syed M. Minhaj Naqvi, Khaled M. Rabie, Abdallah Moubayed arXiv:2610.35873v1 Announce Type: new Abstract: The integration of the Internet of Things (IoT) into monitoring systems has become increasingly significant across various fields, including energy management, environmental monitoring, and industrial applications. This paper proposes a smart, real-time, energy-aware monitoring system architecture for IoT using a brokerless Data Distribution Service (DDS) middleware. Environmental and electrical metrics are transmitted with high reliability, achieving an average latency of 1.76 milliseconds and a throughput of 57.1 Mbps. These metrics are processed by a multi-head Long Short-Term Memory network. Evaluated on the empirical TDC2.0 data center dataset, the architecture simultaneously performs three tasks: predicting thermal stress, detecting cooling inefficiency, and forecasting multi-step cooling power demand. The multi task network achieved an F1-score of 0.92 for proactive thermal stress prediction and demonstrated the capability to identify systemic cooling anomalies. Furthermore, the model generated consistent power forecasts across 5-minute, 10-minute, and 30 minute horizons, maintaining an R2 > 0.83. A post-hoc SHAP analysis indicated that the predictive logic aligns with established thermodynamic relationships. Overall, the proposed architecture offers a transparent and low-latency framework for critical facility management.

arXiv:2610.35873v1 Announce Type: new Abstract: The integration of the Internet of Things (IoT) into monitoring systems has become increasingly significant across various fields, including energy management, environmental monitoring, and industrial applications. This paper proposes a smart, real-time, energy-aware monitoring system architecture for IoT using a brokerless Data Distribution Service (DDS) middleware. Environmental and electrical metrics are transmitted with high reliability, achieving an average latency of 1.76 milliseconds and a throughput of 57.1 Mbps. These metrics are processed by a multi-head Long Short-Term Memory network. Evaluated on the empirical TDC2.0 data center dataset, the architecture simultaneously performs three tasks: predicting thermal stress, detecting cooling inefficiency, and forecasting multi-step cooling power demand. The multi task network achieved an F1-score of 0.92 for proactive thermal stress prediction and demonstrated the capability to identify systemic cooling anomalies. Furthermore, the model generated consistent power forecasts across 5-minute, 10-minute, and 30 minute horizons, maintaining an R2 > 0.83. A post-hoc SHAP analysis indicated that the predictive logic aligns with established thermodynamic relationships. Overall, the proposed architecture offers a transparent and low-latency framework for critical facility management.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35874v1 2026-10-27T20:00:00+00:00 Kousar Malekinasab, Mostafa Abdollah, Jinwei Zhao, Jianping Pan arXiv:2610.35874v1 Announce Type: new Abstract: Low-Earth-Orbit (LEO) satellite networks are expanding broadband access, but major Real-Time Communication (RTC) platforms, including Google Meet, Zoom, and Microsoft Teams, still rely on service-point selection mechanisms that appear to be largely shaped by terrestrial-network assumptions. As a result, they may not consistently account for LEO-specific dynamics such as satellite topology and Point-of-Presence (PoP) changes, especially during in-flight connectivity. To characterize RTC service-point selection over Starlink, we measure RTC performance using a Canada-wide testbed and more than 24 hours of in-flight experiments. Our results show clear platform differences. Google Meet tends to align more closely with the serving PoP, consistent with broader observed edge coverage. Zoom follows a regional service-point selection pattern, while Microsoft Teams often remains tied to fixed service points. Moreover, we show that in-flight PoP handovers can shift traffic egress points mid-session, and that RTC platforms respond differently to these changes. In some cases, this interaction leads to suboptimal service-point selections and RTT inflation. Overall, our findings suggest that LEO connectivity can expose mismatches between Starlink egress dynamics and RTC service-point selection, motivating further measurement and LEO-optimized designs for service placement and selection.

arXiv:2610.35874v1 Announce Type: new Abstract: Low-Earth-Orbit (LEO) satellite networks are expanding broadband access, but major Real-Time Communication (RTC) platforms, including Google Meet, Zoom, and Microsoft Teams, still rely on service-point selection mechanisms that appear to be largely shaped by terrestrial-network assumptions. As a result, they may not consistently account for LEO-specific dynamics such as satellite topology and Point-of-Presence (PoP) changes, especially during in-flight connectivity. To characterize RTC service-point selection over Starlink, we measure RTC performance using a Canada-wide testbed and more than 24 hours of in-flight experiments. Our results show clear platform differences. Google Meet tends to align more closely with the serving PoP, consistent with broader observed edge coverage. Zoom follows a regional service-point selection pattern, while Microsoft Teams often remains tied to fixed service points. Moreover, we show that in-flight PoP handovers can shift traffic egress points mid-session, and that RTC platforms respond differently to these changes. In some cases, this interaction leads to suboptimal service-point selections and RTT inflation. Overall, our findings suggest that LEO connectivity can expose mismatches between Starlink egress dynamics and RTC service-point selection, motivating further measurement and LEO-optimized designs for service placement and selection.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35877v1 2026-10-27T20:00:00+00:00 Ayush Mittal, Dhruv Gupta arXiv:2610.35877v1 Announce Type: new Abstract: Changing a deployment policy can invalidate a world model's uncertainty while leaving the physical transition mechanism and predictor unchanged. For one-step prediction, if source and target laws share the conditional mechanism $P(Y|X)=Q(Y|X)$, then the intervention is identified by the pre-outcome change of measure $w(X)=dQ_X/dP_X$. For policy-controlled dynamics this ratio factors into state-occupancy and action-propensity terms. Weighted conformal calibration transports marginal coverage to the target law; with an estimated ratio $\hat w$, $P_Q\{Y \in C_{\hat w}(X)\} \ge 1-\alpha-TV(Q,\hat Q) \ge 1-\alpha-E_P|w-\hat w|$. Population log-loss regret for a source-versus-target domain classifier bounds the same ratio error. If $h$ is its clipped target posterior, $\rho$ is the target-domain prior, and $\Delta(h)$ is population log-loss regret, then $E_P|w-\hat w_h| \le (1/(\rho\gamma))\sqrt{\Delta(h)/2}$. An exact cancellation improves the dependence from a generic odds-map bound of order $\gamma^{-2}$ to $(\rho\gamma)^{-1}$. Our experiments use a released DINO-feature world model and physical Franka trajectories. Under the strongest controlled state-action shift, ordinary 90% conformal sets cover 55.0%, whereas oracle and learned-ratio calibration cover 91.0% and 90.9%. The learned surrogate remains within 0.9% total variation of the target law. Calibration is relative to the deployment law: the model can remain fixed while calibration follows the intervention. We machine-check in Lean 4 the finite-support algebraic and causal identities used in the classifier-regret argument, including its coverage-composition step.

arXiv:2610.35877v1 Announce Type: new Abstract: Changing a deployment policy can invalidate a world model's uncertainty while leaving the physical transition mechanism and predictor unchanged. For one-step prediction, if source and target laws share the conditional mechanism $P(Y|X)=Q(Y|X)$, then the intervention is identified by the pre-outcome change of measure $w(X)=dQ_X/dP_X$. For policy-controlled dynamics this ratio factors into state-occupancy and action-propensity terms. Weighted conformal calibration transports marginal coverage to the target law; with an estimated ratio $\hat w$, $P_Q\{Y \in C_{\hat w}(X)\} \ge 1-\alpha-TV(Q,\hat Q) \ge 1-\alpha-E_P|w-\hat w|$. Population log-loss regret for a source-versus-target domain classifier bounds the same ratio error. If $h$ is its clipped target posterior, $\rho$ is the target-domain prior, and $\Delta(h)$ is population log-loss regret, then $E_P|w-\hat w_h| \le (1/(\rho\gamma))\sqrt{\Delta(h)/2}$. An exact cancellation improves the dependence from a generic odds-map bound of order $\gamma^{-2}$ to $(\rho\gamma)^{-1}$. Our experiments use a released DINO-feature world model and physical Franka trajectories. Under the strongest controlled state-action shift, ordinary 90% conformal sets cover 55.0%, whereas oracle and learned-ratio calibration cover 91.0% and 90.9%. The learned surrogate remains within 0.9% total variation of the target law. Calibration is relative to the deployment law: the model can remain fixed while calibration follows the intervention. We machine-check in Lean 4 the finite-support algebraic and causal identities used in the classifier-regret argument, including its coverage-composition step.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35902v1 2026-10-27T20:00:00+00:00 Vladimir Yugay, Denis Rozumny, Theo Gevers, Elias Vansteenkiste, Martin R. Oswald, Diogo C. Luvizon arXiv:2610.35902v1 Announce Type: new Abstract: We present StereoSplat, a feed-forward 3D Gaussian Splatting architecture designed explicitly for stereo videos to achieve reliable, metric-scale novel-view synthesis. Unlike monocular approaches that neglect the nature of binocular pairs, StereoSplat adopts a native stereo architecture that extracts dense multi-scale latent features and disparity from a foundation stereo backbone and feeds them into a dedicated Gaussian decoder. Our decoder utilizes cross-view feature fusion and gated depth refinement to directly translate these rich representations into per-pixel 3D Gaussians. To resolve the high redundancy of per-pixel predictions across overlapping views, we propose a learnable depth-adaptive aggregation mechanism that clusters and aggregates primitives in inverse-depth space while preserving fine-grained details. To support training and generalization, we introduce XRStereo, a large-scale synthetic dataset of stereo video featuring various rig configurations tailored to match the physical hardware of real-world platforms. Trained exclusively on synthetic data and evaluated across challenging benchmarks, StereoSplat yields substantial improvements in the depth accuracy of rendered novel views over established baselines while maintaining high-fidelity novel view synthesis. The model also transfers zero-shot to real-world stereo video, with applications in mixed reality, autonomous driving, and robotics.

arXiv:2610.35902v1 Announce Type: new Abstract: We present StereoSplat, a feed-forward 3D Gaussian Splatting architecture designed explicitly for stereo videos to achieve reliable, metric-scale novel-view synthesis. Unlike monocular approaches that neglect the nature of binocular pairs, StereoSplat adopts a native stereo architecture that extracts dense multi-scale latent features and disparity from a foundation stereo backbone and feeds them into a dedicated Gaussian decoder. Our decoder utilizes cross-view feature fusion and gated depth refinement to directly translate these rich representations into per-pixel 3D Gaussians. To resolve the high redundancy of per-pixel predictions across overlapping views, we propose a learnable depth-adaptive aggregation mechanism that clusters and aggregates primitives in inverse-depth space while preserving fine-grained details. To support training and generalization, we introduce XRStereo, a large-scale synthetic dataset of stereo video featuring various rig configurations tailored to match the physical hardware of real-world platforms. Trained exclusively on synthetic data and evaluated across challenging benchmarks, StereoSplat yields substantial improvements in the depth accuracy of rendered novel views over established baselines while maintaining high-fidelity novel view synthesis. The model also transfers zero-shot to real-world stereo video, with applications in mixed reality, autonomous driving, and robotics.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35904v1 2026-10-27T20:00:00+00:00 Zhenghe Guo arXiv:2610.35904v1 Announce Type: new Abstract: Object-wise reconstruction increasingly inserts generated meshes into captured scenes as editable proxies for observed objects. This use creates a requirement absent from isolated-asset scores and pose estimates: at the captured metric state, the proxy should reproduce calibrated observations of the same instance. We present ObjectTwin, a scan-grounded measurement protocol that freezes the instance, cameras, scan-projected support and recorded depth, metric anchor, and evaluation denominator, then replaces only the object representation. The resulting asset correspondence gap measures lost observation agreement relative to scan-derived visible geometry without treating the scan as complete physical ground truth. Across two indoor RGB-D capture domains and two generation backends, visible, anchored generated assets lose 0.302-0.497 MaskIoU under this replacement. Every tested scene-level estimate is positive, while same-input generator preference reverses across capture domains. Coverage and overlap-depth analyses locate disagreement in both visible support and surface placement; the evaluated conservative pose refinement does not systematically recover it. These results establish capture correspondence as a separately measurable interface between generated-asset quality and stronger physical or task-level evaluation.

arXiv:2610.35904v1 Announce Type: new Abstract: Object-wise reconstruction increasingly inserts generated meshes into captured scenes as editable proxies for observed objects. This use creates a requirement absent from isolated-asset scores and pose estimates: at the captured metric state, the proxy should reproduce calibrated observations of the same instance. We present ObjectTwin, a scan-grounded measurement protocol that freezes the instance, cameras, scan-projected support and recorded depth, metric anchor, and evaluation denominator, then replaces only the object representation. The resulting asset correspondence gap measures lost observation agreement relative to scan-derived visible geometry without treating the scan as complete physical ground truth. Across two indoor RGB-D capture domains and two generation backends, visible, anchored generated assets lose 0.302-0.497 MaskIoU under this replacement. Every tested scene-level estimate is positive, while same-input generator preference reverses across capture domains. Coverage and overlap-depth analyses locate disagreement in both visible support and surface placement; the evaluated conservative pose refinement does not systematically recover it. These results establish capture correspondence as a separately measurable interface between generated-asset quality and stronger physical or task-level evaluation.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35905v1 2026-10-27T20:00:00+00:00 Jirong Zha, Yanwei Liao, Yemin Wang, Xiangmin Yi, Siqi Peng, Yingfeng Chen, Chen Gao, Xinlei Chen arXiv:2610.35905v1 Announce Type: new Abstract: Image quality assessment (IQA) underpins embodied imaging pipelines by judging whether visual quality satisfies downstream tasks, yet most existing methods learn task-agnostic scores aligned with generic human ratings on static benchmarks. This objective mismatches embodied and interactive settings, where image adequacy depends on task goals, context, and action requirements that shape an agent's decisions. We argue that IQA should shift from score regression to goal-conditioned judgment defined by embodied task utility. To this end, we propose an MLLM-based embodied IQA agent framework that reasons about task requirements, grounds quality evidence, and produces decision-facing assessments. A compact VisDrone evidence study shows that conventional perceptual metrics and generic MLLM-based IQA scores are weak proxies for downstream VLM task success, while a task-conditioned utility score better predicts whether visual input supports the target task. These findings support a task-utility-oriented reframing of IQA for embodied, explainable, and decision-aware visual perception.

arXiv:2610.35905v1 Announce Type: new Abstract: Image quality assessment (IQA) underpins embodied imaging pipelines by judging whether visual quality satisfies downstream tasks, yet most existing methods learn task-agnostic scores aligned with generic human ratings on static benchmarks. This objective mismatches embodied and interactive settings, where image adequacy depends on task goals, context, and action requirements that shape an agent's decisions. We argue that IQA should shift from score regression to goal-conditioned judgment defined by embodied task utility. To this end, we propose an MLLM-based embodied IQA agent framework that reasons about task requirements, grounds quality evidence, and produces decision-facing assessments. A compact VisDrone evidence study shows that conventional perceptual metrics and generic MLLM-based IQA scores are weak proxies for downstream VLM task success, while a task-conditioned utility score better predicts whether visual input supports the target task. These findings support a task-utility-oriented reframing of IQA for embodied, explainable, and decision-aware visual perception.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35906v1 2026-10-27T20:00:00+00:00 Yinyi Luo, Hayes Bai, Yiqiao Jin, Wenyue Hua, Srijan Kumar, Marios Savvides, Qingsong Wen, Jindong Wang arXiv:2610.35906v1 Announce Type: new Abstract: Unified Multimodal Models (UMMs) emerge as a promising way to combine understanding and generation within a single architecture. We argue that current approaches mainly achieve structural unification, which can create a Unification Paradox: shared interfaces, token spaces, and parameters do not guarantee an overall capability gain and may degrade individual capabilities. In this position paper, we identify five critical layers in which unification must be rethought: (1) Asymmetric Input Unification, where models must estimate the utility and reliability of each modality rather than treating all modalities equally; (2) Factorized Representation Unification, where semantic abstraction and fine-grained reconstruction should use separate but coordinated capacity; (3) Stateful Reasoning Unification, where models should maintain and revise a shared multimodal belief state across understanding and generation; (4) Synergistic Optimization Unification, where conflicting understanding and generation objectives require coordinated optimization; and (5) Multi-Centric Systemic Evaluation, where benchmarks must evaluate systemic competence, capability preservation, and cross-task robustness rather than isolated outcomes. We argue that progress requires moving beyond structural unification toward functional unification across the full system.

arXiv:2610.35906v1 Announce Type: new Abstract: Unified Multimodal Models (UMMs) emerge as a promising way to combine understanding and generation within a single architecture. We argue that current approaches mainly achieve structural unification, which can create a Unification Paradox: shared interfaces, token spaces, and parameters do not guarantee an overall capability gain and may degrade individual capabilities. In this position paper, we identify five critical layers in which unification must be rethought: (1) Asymmetric Input Unification, where models must estimate the utility and reliability of each modality rather than treating all modalities equally; (2) Factorized Representation Unification, where semantic abstraction and fine-grained reconstruction should use separate but coordinated capacity; (3) Stateful Reasoning Unification, where models should maintain and revise a shared multimodal belief state across understanding and generation; (4) Synergistic Optimization Unification, where conflicting understanding and generation objectives require coordinated optimization; and (5) Multi-Centric Systemic Evaluation, where benchmarks must evaluate systemic competence, capability preservation, and cross-task robustness rather than isolated outcomes. We argue that progress requires moving beyond structural unification toward functional unification across the full system.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35908v1 2026-10-27T20:00:00+00:00 Yuyang Shi arXiv:2610.35908v1 Announce Type: new Abstract: Emoji can easily lead to misunderstanding. On WeChat, [Smile] is often used by the elderly to express positive emotions such as joy and praise, but young people often regard [Smile] as sarcasm. In this case, misunderstandings can easily occur. In order to confirm the existence of this phenomenon and analyze people's understanding of emoji, this study selected 20 smiley emoji on WeChat platform and directly asked people's opinions on these emoji through questionnaires, including whether they recommend others to use, and rating of emotion joy, praise, sarcasm. Analyses were performed based on the total sample and by age group. The results showed that [Smile] and [Bye] were interpreted in opposite ways by older people and younger people. Older people thought they represented positive emotion (joy, praise), while younger people thought they represented sarcasm. Such emojis are easy to misunderstand. On the other hand, some emojis like [Yeah!], both the old and the young have a similar understanding that it represents a positive emotion.

arXiv:2610.35908v1 Announce Type: new Abstract: Emoji can easily lead to misunderstanding. On WeChat, [Smile] is often used by the elderly to express positive emotions such as joy and praise, but young people often regard [Smile] as sarcasm. In this case, misunderstandings can easily occur. In order to confirm the existence of this phenomenon and analyze people's understanding of emoji, this study selected 20 smiley emoji on WeChat platform and directly asked people's opinions on these emoji through questionnaires, including whether they recommend others to use, and rating of emotion joy, praise, sarcasm. Analyses were performed based on the total sample and by age group. The results showed that [Smile] and [Bye] were interpreted in opposite ways by older people and younger people. Older people thought they represented positive emotion (joy, praise), while younger people thought they represented sarcasm. Such emojis are easy to misunderstand. On the other hand, some emojis like [Yeah!], both the old and the young have a similar understanding that it represents a positive emotion.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35910v1 2026-10-27T20:00:00+00:00 Soumyajit Datta, Arka Dutta, Ashiqur R. KhudaBukhsh arXiv:2610.35910v1 Announce Type: new Abstract: Emergent generative AI technologies, while remarkably beneficial across several tasks and domains, can suffer from unforeseen vulnerabilities and difficult-to-anticipate failure modes. Such failure modes can disparately affect vulnerable communities. This paper investigates two well-known vision language models (VLMs) and inspects how these models encode protected identities while generating images. We show that when VLMs are requested to transform human face images to reflect race identities (e.g., Hispanic, Black, White, or Asian) and sexual orientation (e.g., gay or straight), VLMs often comply with such requests. We demonstrate that compliance (and refusal) is non-deterministic, group-sensitive, and generated images are statistically distinguishable, often accentuating deep-seated stereotypes. We further show that representation biases introduced during image generation can re-emerge in textual reasoning and characterization tasks. Finally, we show that this worrisome finding is not limited to racial identity or sexual orientation, image generation can perpetuate bias around criminality, raising concerning possibilities of AI-assisted profiling.

arXiv:2610.35910v1 Announce Type: new Abstract: Emergent generative AI technologies, while remarkably beneficial across several tasks and domains, can suffer from unforeseen vulnerabilities and difficult-to-anticipate failure modes. Such failure modes can disparately affect vulnerable communities. This paper investigates two well-known vision language models (VLMs) and inspects how these models encode protected identities while generating images. We show that when VLMs are requested to transform human face images to reflect race identities (e.g., Hispanic, Black, White, or Asian) and sexual orientation (e.g., gay or straight), VLMs often comply with such requests. We demonstrate that compliance (and refusal) is non-deterministic, group-sensitive, and generated images are statistically distinguishable, often accentuating deep-seated stereotypes. We further show that representation biases introduced during image generation can re-emerge in textual reasoning and characterization tasks. Finally, we show that this worrisome finding is not limited to racial identity or sexual orientation, image generation can perpetuate bias around criminality, raising concerning possibilities of AI-assisted profiling.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35925v1 2026-10-27T20:00:00+00:00 Taehwan Lee arXiv:2610.35925v1 Announce Type: new Abstract: Training contextual masking systems for personally identifiable information (PII) requires examples distinguishing maskable from preservable mentions of one entity type, but privacy restrictions limit access to real data. Author-based augmentation (SPY) labels mentions unrelated to the author as non-PII, presuming a single data subject and overlooking others. We propose ReCoPII (Relation-grounded Contextual PII masking), a framework that builds such data without this premise: it plans ontology-constrained entity relations, writes insertion sentences with placeholders, reconstructs annotations with code, checks the result with a separate LLM, and derives MASK/KEEP labels from entity relations. On Korean documents from six domains, it yields ReCoPII-Ko, a corpus with relation, privacy, and MASK/KEEP annotations, which we validate. Across 300 pairs, four annotators agree with 88.42% of its sampled KEEP labels, against 26.27% for SPY-style augmentation of the same sources. Classifiers trained on it over-mask less than under that reference on a matched test. Contextual masking fine-tuned jointly with a PII detector removes most over-masking of detection alone within a fixed leakage budget; on human-verified entity records, the four detectors meeting it over-mask at most 5.04% with at most 4.18% leakage. The detector is comparable to XLM-R-large on KDPII with about a quarter of its parameters.

arXiv:2610.35925v1 Announce Type: new Abstract: Training contextual masking systems for personally identifiable information (PII) requires examples distinguishing maskable from preservable mentions of one entity type, but privacy restrictions limit access to real data. Author-based augmentation (SPY) labels mentions unrelated to the author as non-PII, presuming a single data subject and overlooking others. We propose ReCoPII (Relation-grounded Contextual PII masking), a framework that builds such data without this premise: it plans ontology-constrained entity relations, writes insertion sentences with placeholders, reconstructs annotations with code, checks the result with a separate LLM, and derives MASK/KEEP labels from entity relations. On Korean documents from six domains, it yields ReCoPII-Ko, a corpus with relation, privacy, and MASK/KEEP annotations, which we validate. Across 300 pairs, four annotators agree with 88.42% of its sampled KEEP labels, against 26.27% for SPY-style augmentation of the same sources. Classifiers trained on it over-mask less than under that reference on a matched test. Contextual masking fine-tuned jointly with a PII detector removes most over-masking of detection alone within a fixed leakage budget; on human-verified entity records, the four detectors meeting it over-mask at most 5.04% with at most 4.18% leakage. The detector is comparable to XLM-R-large on KDPII with about a quarter of its parameters.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35926v1 2026-10-27T20:00:00+00:00 Fabricio Venturim, Leighton Estrada, Claudio Silva, Jorge Poco arXiv:2610.35926v1 Announce Type: new Abstract: Compiler feedback certifies that Lean accepts a declaration, not that the declaration captures the intended result: a repository can compile while a central theorem is weakened, merely auxiliary, or transitively dependent on sorry. Formalization progress should instead be measured against the source paper, in terms of its paper items: definitions, lemmas, and theorems. LeanTrail implements this view as a post-hoc, tool-agnostic audit: deterministic retrieval proposes candidate counterparts for each paper item in the Lean repository and, when available, the Blueprint; a language model judges statement-level correspondence, only among those candidates; and proof completeness comes from a deterministic transitive dependency audit never exposed to the model. Fusion assigns each item one of five coverage states, from Formalized to Missing, each backed by inspectable evidence. On consensus annotations of 98 items from three Lean projects, with three runs of each of two judge models, LeanTrail assigns the correct coverage state to 91.8% (Gemini 3.8 Flash) and 92.5% (GPT-6 Sol) of items. 96-97 of the 98 states are identical across runs, and GPT-6 Sol never reports a Partial item as Formalized. Uncovered items form a natural task queue for formalization agents, anchored to the specification rather than the compiler.

arXiv:2610.35926v1 Announce Type: new Abstract: Compiler feedback certifies that Lean accepts a declaration, not that the declaration captures the intended result: a repository can compile while a central theorem is weakened, merely auxiliary, or transitively dependent on sorry. Formalization progress should instead be measured against the source paper, in terms of its paper items: definitions, lemmas, and theorems. LeanTrail implements this view as a post-hoc, tool-agnostic audit: deterministic retrieval proposes candidate counterparts for each paper item in the Lean repository and, when available, the Blueprint; a language model judges statement-level correspondence, only among those candidates; and proof completeness comes from a deterministic transitive dependency audit never exposed to the model. Fusion assigns each item one of five coverage states, from Formalized to Missing, each backed by inspectable evidence. On consensus annotations of 98 items from three Lean projects, with three runs of each of two judge models, LeanTrail assigns the correct coverage state to 91.8% (Gemini 3.8 Flash) and 92.5% (GPT-6 Sol) of items. 96-97 of the 98 states are identical across runs, and GPT-6 Sol never reports a Partial item as Formalized. Uncovered items form a natural task queue for formalization agents, anchored to the specification rather than the compiler.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35929v1 2026-10-27T20:00:00+00:00 Ilyas Nagiyev, Miras Baisbay, Daniil Orel, Preslav Nakov arXiv:2610.35929v1 Announce Type: new Abstract: We present SCOPE, a multi-agent system for profiling news media outlets with respect to bias, factuality, and credibility. Given an outlet URL, SCOPE coordinates specialized agents for article acquisition, content analysis, external evidence search, scoring, and report generation. The system follows an MBFC-style methodology while making intermediate evidence and agent outputs inspectable through a structured demonstration interface. On a benchmark of 200 media outlets, SCOPE outperforms LLM-only, article-only, search-only, and media-background-check baselines on most metrics, achieving the lowest bias and factuality MAE, perfect failed-fact-check detection, and the lowest factual error rate. These results suggest that decomposed, evidence-grounded agent workflows can support more transparent and reliable media profiling. Our code is available at https://github.com/mbaisbay/SCOPE, and a live demonstration at https://scope-demo.streamlit.app/.

arXiv:2610.35929v1 Announce Type: new Abstract: We present SCOPE, a multi-agent system for profiling news media outlets with respect to bias, factuality, and credibility. Given an outlet URL, SCOPE coordinates specialized agents for article acquisition, content analysis, external evidence search, scoring, and report generation. The system follows an MBFC-style methodology while making intermediate evidence and agent outputs inspectable through a structured demonstration interface. On a benchmark of 200 media outlets, SCOPE outperforms LLM-only, article-only, search-only, and media-background-check baselines on most metrics, achieving the lowest bias and factuality MAE, perfect failed-fact-check detection, and the lowest factual error rate. These results suggest that decomposed, evidence-grounded agent workflows can support more transparent and reliable media profiling. Our code is available at https://github.com/mbaisbay/SCOPE, and a live demonstration at https://scope-demo.streamlit.app/.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35930v1 2026-10-27T20:00:00+00:00 Arnesh Batra, Aniket Khandelwal, Jashn Khemani, Krish Thukral, Arush Gumber, Denis Federico Lim, Ujwal Kumar, Phan Xuan Tan arXiv:2610.35930v1 Announce Type: new Abstract: Multi-agent debate (MAD) aims to improve language-model decisions through the exchange of arguments, yet agreement after debate need not reflect independent corroboration. We investigate how peer communication influences individual decisions and collective outcomes through behavioral and internal interventions. We introduce MAD Bench, a 2,400-question benchmark spanning knowledge and reasoning, values and moral judgment, and safety and refusal, and evaluate three open-weight models using a round-matched independent-reflection control. Among answer-changing responses, a forced-choice readout before rationale generation matches the eventual answer more often in debate than in reflection (65.4% versus 22.3% on Qwen3.5-9B). Message ablations reveal model-dependent effects of peer answers and reasoning, while attributing identical text to the receiver rather than another agent substantially changes its influence. For the tested source descriptions, stated track record affects answer preferences more strongly than professional-status cues across all three models. Attention masking produces different changes in decision margins across the three checkpoints. On fully parsed, reference-labelled non-numeric questions, debate yields small estimated accuracy improvements over initial majority voting, while increasing the rate of unanimous wrong answers by 3.1-7.4 percentage points relative to reflection. Longer debates show persistent consensus and greater correlation between agents' correctness than reflection. These findings show that stronger consensus can coexist with more shared errors, motivating evaluations that assess dependence and unanimous errors alongside accuracy.

arXiv:2610.35930v1 Announce Type: new Abstract: Multi-agent debate (MAD) aims to improve language-model decisions through the exchange of arguments, yet agreement after debate need not reflect independent corroboration. We investigate how peer communication influences individual decisions and collective outcomes through behavioral and internal interventions. We introduce MAD Bench, a 2,400-question benchmark spanning knowledge and reasoning, values and moral judgment, and safety and refusal, and evaluate three open-weight models using a round-matched independent-reflection control. Among answer-changing responses, a forced-choice readout before rationale generation matches the eventual answer more often in debate than in reflection (65.4% versus 22.3% on Qwen3.5-9B). Message ablations reveal model-dependent effects of peer answers and reasoning, while attributing identical text to the receiver rather than another agent substantially changes its influence. For the tested source descriptions, stated track record affects answer preferences more strongly than professional-status cues across all three models. Attention masking produces different changes in decision margins across the three checkpoints. On fully parsed, reference-labelled non-numeric questions, debate yields small estimated accuracy improvements over initial majority voting, while increasing the rate of unanimous wrong answers by 3.1-7.4 percentage points relative to reflection. Longer debates show persistent consensus and greater correlation between agents' correctness than reflection. These findings show that stronger consensus can coexist with more shared errors, motivating evaluations that assess dependence and unanimous errors alongside accuracy.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35933v1 2026-10-27T20:00:00+00:00 Rama Cont, Massoud Heidari arXiv:2610.35933v1 Announce Type: new Abstract: We study deterministic causal quantization with pathwise error at most $\varepsilon$, using $\omega$ shared representatives updated from transmitted addresses alone. For a midpoint realization with first-in, first-out replacement, we compare address entropy rates with process $\varepsilon$-entropy, the least entropy rate of any block quantizer meeting the same constraint. From $\varepsilon$-separated initialization, the gap is at most two bits per sample for every stationary source with finite scalar $\varepsilon$-entropy, without density, independence or mixing assumptions. A causal address-tree argument extends this comparison to any synchronized quantizer with cell overlap $m$, giving $\log_2 m$ bits. Lattice initialization with consistent boundary ownership realizes uniform grids with sharp one-bit scalar and process comparisons. Under an entropy condition and finite Fisher information rate $J$, the gap is at most $2J\varepsilon^2/(\pi^2\ln 2)$; general initializations add a term of order $\omega\varepsilon\sqrt{J}$. The scalar grid estimate gives a nonasymptotic form of R\'enyi's theorem. Since a grid is causal, the price of causality is at most $\min\{1,2J\varepsilon^2/(\pi^2\ln 2)\}$ bits per sample, versus the high-resolution squared-error loss $\tfrac12\log_2(\pi e/6)$. These bounds certify $\varepsilon$-entropy brackets for autoregressive Laplace and Gaussian sources. For memoryless sources, the optimal causal rate equals the scalar $\varepsilon$-entropy. For a stationary process on finitely many $2\varepsilon$-separated centers, adding independent uniform noise of width $2\varepsilon$ gives exact Shannon lower-bound equality at every block length, attained by storing those centers. This includes histograms with finitely many bins. Storing two non-lattice centers beats every uniform grid of the same cell width by half a bit per sample.

arXiv:2610.35933v1 Announce Type: new Abstract: We study deterministic causal quantization with pathwise error at most $\varepsilon$, using $\omega$ shared representatives updated from transmitted addresses alone. For a midpoint realization with first-in, first-out replacement, we compare address entropy rates with process $\varepsilon$-entropy, the least entropy rate of any block quantizer meeting the same constraint. From $\varepsilon$-separated initialization, the gap is at most two bits per sample for every stationary source with finite scalar $\varepsilon$-entropy, without density, independence or mixing assumptions. A causal address-tree argument extends this comparison to any synchronized quantizer with cell overlap $m$, giving $\log_2 m$ bits. Lattice initialization with consistent boundary ownership realizes uniform grids with sharp one-bit scalar and process comparisons. Under an entropy condition and finite Fisher information rate $J$, the gap is at most $2J\varepsilon^2/(\pi^2\ln 2)$; general initializations add a term of order $\omega\varepsilon\sqrt{J}$. The scalar grid estimate gives a nonasymptotic form of R\'enyi's theorem. Since a grid is causal, the price of causality is at most $\min\{1,2J\varepsilon^2/(\pi^2\ln 2)\}$ bits per sample, versus the high-resolution squared-error loss $\tfrac12\log_2(\pi e/6)$. These bounds certify $\varepsilon$-entropy brackets for autoregressive Laplace and Gaussian sources. For memoryless sources, the optimal causal rate equals the scalar $\varepsilon$-entropy. For a stationary process on finitely many $2\varepsilon$-separated centers, adding independent uniform noise of width $2\varepsilon$ gives exact Shannon lower-bound equality at every block length, attained by storing those centers. This includes histograms with finitely many bins. Storing two non-lattice centers beats every uniform grid of the same cell width by half a bit per sample.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35935v1 2026-10-27T20:00:00+00:00 Ayushman Bhattacharya, Anwesha Chakraborty arXiv:2610.35935v1 Announce Type: new Abstract: Video on SPI-attached microcontroller displays is limited less by decoding than by where each byte lives and which stages run serially. We present RV565, a decode framework for frame-independent video whose decoded representation is the panel's native RGB565 format, and evaluate it on an ESP32-S3 with two ten-second windows of an openly licensed film. A serial path - inflate, then scale into internal-memory stripes overlapped with SPI direct-memory-access (DMA) transfers - meets the 41.7 ms deadline on the lighter window but misses all 7,200 frames of the heavier one. RV565's dual-core pipeline inflates the next frame on the second core during the transfer and sustains 24 FPS with zero misses and 10.3 ms of slack on both windows; frame time then equals the display transfer, which runs at 98.1% of the 40 MHz wire-time bound. Because frames are independent, inflation time is linear in the compressed length stored in the container (pooled r^2 = 0.9987): a model fitted on one window predicts the other within 0.04 ms and identifies from metadata alone which window the serial path cannot schedule. We show why Deflate cannot exploit co-located temporal redundancy in RGB565 frames larger than 128x128 pixels and measure the price: a shared Deflate stream changes size by at most 1.1%, whereas an explicit XOR delta would save 46-80%. Inside OreoOS, a MicroPython application operating system, firmware inflation alone reaches 12.2 FPS; stage timing attributes the 80.4 ms frame time to serialized filesystem input, external-RAM placement, and blocking presentation, with 0.8 ms of interpreter overhead. Firmware, measurements, and analysis scripts are released.

arXiv:2610.35935v1 Announce Type: new Abstract: Video on SPI-attached microcontroller displays is limited less by decoding than by where each byte lives and which stages run serially. We present RV565, a decode framework for frame-independent video whose decoded representation is the panel's native RGB565 format, and evaluate it on an ESP32-S3 with two ten-second windows of an openly licensed film. A serial path - inflate, then scale into internal-memory stripes overlapped with SPI direct-memory-access (DMA) transfers - meets the 41.7 ms deadline on the lighter window but misses all 7,200 frames of the heavier one. RV565's dual-core pipeline inflates the next frame on the second core during the transfer and sustains 24 FPS with zero misses and 10.3 ms of slack on both windows; frame time then equals the display transfer, which runs at 98.1% of the 40 MHz wire-time bound. Because frames are independent, inflation time is linear in the compressed length stored in the container (pooled r^2 = 0.9987): a model fitted on one window predicts the other within 0.04 ms and identifies from metadata alone which window the serial path cannot schedule. We show why Deflate cannot exploit co-located temporal redundancy in RGB565 frames larger than 128x128 pixels and measure the price: a shared Deflate stream changes size by at most 1.1%, whereas an explicit XOR delta would save 46-80%. Inside OreoOS, a MicroPython application operating system, firmware inflation alone reaches 12.2 FPS; stage timing attributes the 80.4 ms frame time to serialized filesystem input, external-RAM placement, and blocking presentation, with 0.8 ms of interpreter overhead. Firmware, measurements, and analysis scripts are released.

2026-10-27T20:00:00+00:00 oai:arXiv.org:2610.35942v1 2026-10-27T20:00:00+00:00 Solomon Shalom Lijo arXiv:2610.35942v1 Announce Type: new Abstract: I consider what it means for AI assistance to save time. I contrast a stopwatch view, on which time saved is whatever the clock records, with a ledger view, on which it is a counterfactual that has to be read next to the errors the work lets through, and I argue for the latter. The argument has four parts. I define a model-free pair of quantities, time saved and excess escaped defects, and one decision number built from them: the break-even defect cost. I give a baseline model of checking in which, until the model is good enough not to check, a calibrated worker's escaped-defect rate does not depend on how good the model is; checking time grows with the logarithm of the model's error rate; and over-trust buys minutes on the clock at a larger cost in total minutes. I show what randomized studies can and cannot identify, and I replicate METR's developer study from its public data: the same 246 issues support a slowdown of 19% or 24% depending on the estimand (34% without adjusting for a chance imbalance), and AI helped on somewhere between 3% and 82% of them. Finally, I restate the published record, with stated bounds in place of imported point assumptions. In GDPval's own review-every-output workflow, review time alone caps GPT-5's speedup at 3.6x; the consultants outside the frontier were faster, and more often wrong. Last, I test two of the model's predictions in a small pre-specified experiment on AI reviewers checking AI drafts of the same kind of brief, built from SEC filings made after the models' training data were collected. Told the drafter's true error rate, the reviewer checked more as it rose, but all-or-nothing, switching near where its prompt's costs break even. Told that the drafter erred ten times less often than it did, it checked 7 fewer of 9 figures and let a wrong figure through in 22.5% of briefs instead of none: faster and worse, entirely through reduced effort.

arXiv:2610.35942v1 Announce Type: new Abstract: I consider what it means for AI assistance to save time. I contrast a stopwatch view, on which time saved is whatever the clock records, with a ledger view, on which it is a counterfactual that has to be read next to the errors the work lets through, and I argue for the latter. The argument has four parts. I define a model-free pair of quantities, time saved and excess escaped defects, and one decision number built from them: the break-even defect cost. I give a baseline model of checking in which, until the model is good enough not to check, a calibrated worker's escaped-defect rate does not depend on how good the model is; checking time grows with the logarithm of the model's error rate; and over-trust buys minutes on the clock at a larger cost in total minutes. I show what randomized studies can and cannot identify, and I replicate METR's developer study from its public data: the same 246 issues support a slowdown of 19% or 24% depending on the estimand (34% without adjusting for a chance imbalance), and AI helped on somewhere between 3% and 82% of them. Finally, I restate the published record, with stated bounds in place of imported point assumptions. In GDPval's own review-every-output workflow, review time alone caps GPT-5's speedup at 3.6x; the consultants outside the frontier were faster, and more often wrong. Last, I test two of the model's predictions in a small pre-specified experiment on AI reviewers checking AI drafts of the same kind of brief, built from SEC filings made after the models' training data were collected. Told the drafter's true error rate, the reviewer checked more as it rose, but all-or-nothing, switching near where its prompt's costs break even. Told that the drafter erred ten times less often than it did, it checked 7 fewer of 9 figures and let a wrong figure through in 22.5% of briefs instead of none: faster and worse, entirely through reduced effort.

2026-10-27T20:00:00+00:00