arXiv subjects

Shashank Mishra

Publications and source records attributed to Shashank Mishra.

15 recordsLinked to original sources

arXiv2026

When Does Memory Help? A Cost-Aware Evaluation of Long-Term Memory in Tool-Using LLM Agents

Long-term memory for LLM agents is evaluated today by conversational recall benchmarks (LoCoMo, LongMemEval), which measure question answering over dialogue history, not whether remembered facts change what a tool-using agent does. We present MERIT (Memory Evaluation for Realistic Instrumented Tasks), a benchmark and harness that measures the marginal utility of memory for task-executing agents under explicit cost accounting. MERIT provides episodic tool-use tasks in three domains whose dependence on earlier-episode facts is verified by an automated leak check; a difficulty ladder ending in updated-fact recall; controlled memory corruption; and full token and dollar metering of every memory operation. Across 23,440 scored episodes ($42.57), a two-generation pilot on gpt-4.1-mini and a preregistered 3-model x 3-seed grid (GPT-4.1, Claude Haiku 4.5; memory side held fixed), memory lifts dependent-task success from a leak-verified floor of 0.00 to 0.55-1.00. On updated facts, embedding retrieval collapses unpredictably (0.30-0.95 across models; max seed gap 0.45), and agents act on a correctly retrieved value only 55% of the time, while update-on-write stores (a structured fact store and, notably, LLM summarization) remain at 0.70-1.00; the hybrid is worse than the fact store alone. A latest-generation spot-check (Claude Sonnet 5, gated on a clean full-replay control) reproduces the pattern. Swapping a memory's implementation moves task success by up to 60 points, and full replay is never economical: the best condition per domain delivers 2.7-3.9x its marginal utility per dollar. We release the benchmark, harness, and all traces.

cs.AI↗

arXiv2026

Conditional Attribution for Root Cause Analysis in Time-Series Anomaly Detection

Root cause analysis (RCA) for time-series anomaly detection is critical for the reliable operation of complex real-world systems. Existing explanation methods often rely on unrealistic feature perturbations and ignore temporal and cross-feature dependencies, leading to unreliable attributions. We propose a conditional attribution framework that explains anomalies relative to contextually similar normal system states. Instead of using marginal or randomly sampled baselines, our method retrieves representative normal instances conditioned on the anomalous observation, enabling dependency-preserving and operationally meaningful explanations. To support high-dimensional time-series data, contextual retrieval is performed in learned low-dimensional representations using both variational autoencoder latent spaces and UMAP manifold embeddings. By grounding the retrieval process in the system's learned manifold, this strategy avoids out-of-distribution artifacts and ensures attribution fidelity while maintaining computational efficiency. We further introduce confidence-aware and temporal evaluation metrics for assessing explanation reliability and responsiveness. Experiments on the SWaT and MSDS benchmarks demonstrate that the proposed approach consistently improves root-cause identification accuracy, temporal localization, and robustness across multiple anomaly detection models. These results highlight the practical utility of conditional attribution for explainable anomaly diagnosis in complex time-series systems. Code and models are available at: https://github.com/dfki-av/Conditional-Attribution-for-Root-Cause-Analysis-in-Time-Series-Anomaly-Detection.

cs.LG↗

arXiv2026

Conditional Compatibility Learning for Context-Dependent Anomaly Detection

Anomaly detection usually assumes that abnormality is an intrinsic property of an observation. A defect is a defect, and a rare object is rare, regardless of where it appears. Many real-world anomalies do not work this way. A runner on a track is normal, but the same runner on a highway is not. The subject is unchanged; only the context makes it anomalous. This setting, long recognized as contextual anomaly detection, remains largely underexplored in modern vision-language systems. The difficulty is not merely empirical; it is formal. When anomaly labels depend on the relation between a subject and its context, any detector reasoning from a global representation that conflates subject and context is provably non-identifiable: two different subject-context configurations can map to the same embedding while requiring opposite labels, and no such detector can be correct on both. This impossibility motivates a different formulation: instead of asking whether an observation deviates from a global notion of normality, the model should ask whether subjects are compatible with their surrounding context. We define this as conditional compatibility learning. We instantiate this framework in CC-CLIP, a vision-language architecture that learns disentangled subject- and context-aware representations from a single image and fuses visual evidence through text-conditioned attention. CC-CLIP achieves state-of-the-art results on real-world contextual anomaly detection, substantially outperforming all existing CLIP-based and context-reasoning baselines. A single-branch variant of CC-CLIP also achieves competitive performance on structural anomaly benchmarks.

cs.CV↗

arXiv2025

Coexistence of inequivalent time-crystalline orders in a Floquet collective spin system

We investigate the dynamical phases that emerge in collective spin models subjected to a spatially non-uniform periodic drive. Taking the paradigmatic Lipkin-Meshkov-Glick (LMG) model as a concrete platform, we establish that a rich landscape of dynamical phases emerges when two regions of the system are driven with different field strengths, $h_1$ and $h_2$. Remarkably, despite the `all-to-all' nature of the interactions, the system can be driven into dynamical phases characterized by distinct kinds of discrete time crystal (DTC) orders in different parts of the system. Apart from these coexisting DTCs, tuning the driving field leads to the emergence of phases where DTCs coexist with Floquet-synchronized or oscillatory phases; the former has been dubbed a chimera DTC. Finally, we demonstrate that a tunable set of global DTC phases emerges when $h_1$ and $h_2$ are proximate. Crucially, these dynamical regimes can be observed both for experimentally relevant finite-size systems and in the thermodynamic limit. Our results establish spatially structured driving as a powerful route to realize non-equilibrium phase coexistence in collective spin systems.

quant-ph↗

arXiv2025

IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusion

High-performance Radar-Camera 3D object detection can be achieved by leveraging knowledge distillation without using LiDAR at inference time. However, existing distillation methods typically transfer modality-specific features directly to each sensor, which can distort their unique characteristics and degrade their individual strengths. To address this, we introduce IMKD, a radar-camera fusion framework based on multi-level knowledge distillation that preserves each sensor's intrinsic characteristics while amplifying their complementary strengths. IMKD applies a three-stage, intensity-aware distillation strategy to enrich the fused representation across the architecture: (1) LiDAR-to-Radar intensity-aware feature distillation to enhance radar representations with fine-grained structural cues, (2) LiDAR-to-Fused feature intensity-guided distillation to selectively highlight useful geometry and depth information at the fusion level, fostering complementarity between the modalities rather than forcing them to align, and (3) Camera-Radar intensity-guided fusion mechanism that facilitates effective feature alignment and calibration. Extensive experiments on the nuScenes benchmark show that IMKD reaches 67.0% NDS and 61.0% mAP, outperforming all prior distillation-based radar-camera fusion methods. Our code and models are available at https://github.com/dfki-av/IMKD/.

cs.CV↗

arXiv2025

A Multi-Camera Vision-Based Approach for Fine-Grained Assembly Quality Control

Quality control is a critical aspect of manufacturing, particularly in ensuring the proper assembly of small components in production lines. Existing solutions often rely on single-view imaging or manual inspection, which are prone to errors due to occlusions, restricted perspectives, or lighting inconsistencies. These limitations require the installation of additional inspection stations, which could disrupt the assembly line and lead to increased downtime and costs. This paper introduces a novel multi-view quality control module designed to address these challenges, integrating a multi-camera imaging system with advanced object detection algorithms. By capturing images from three camera views, the system provides comprehensive visual coverage of components of an assembly process. A tailored image fusion methodology combines results from multiple views, effectively resolving ambiguities and enhancing detection reliability. To support this system, we developed a unique dataset comprising annotated images across diverse scenarios, including varied lighting conditions, occlusions, and angles, to enhance applicability in real-world manufacturing environments. Experimental results show that our approach significantly outperforms single-view methods, achieving high precision and recall rates in the identification of improperly fastened small assembly parts such as screws. This work contributes to industrial automation by overcoming single-view limitations, and providing a scalable, cost-effective, and accurate quality control mechanism that ensures the reliability and safety of the assembly line. The dataset used in this study is publicly available to facilitate further research in this domain.

cs.CV↗

arXiv2024

Exploring Non-Isotropic Lorentz Invariance Violation Through Sidereal Effect at DUNE

Lorentz Invariance Violation (LIV) presents an intriguing opportunity to investigate fundamental symmetries, with neutrinos serving as a particularly effective probe for this phenomenon. Long-baseline neutrino experiments, such as the Deep Underground Neutrino Experiment (DUNE), excel at exploring non-isotropic LIV, especially through the observation of sidereal effects. This study comprehensively examines the full parameter space of non-isotropic, non-diagonal LIV parameters with sidereal dependence, focusing on two distinct flux scenarios: a low-energy flux and a tau-optimized flux. Through this analysis, we derive more stringent constraints on LIV parameters. Our results indicate that DUNE may achieve enhanced sensitivity for some LIV parameters, exceeding all previously established limits and marking a significant advancement in the investigation of LIV.

hep-ph↗

arXiv2024

Octant Ambiguity in the Presence of Non-isotropic Lorentz Invariance Violation

Global analyses of neutrino data suggest that the mixing angle $\theta_{23}$ is likely to be nonmaximal with two closely matched solutions emerging: one representing a smaller angle ($\theta_{23}$ < $\pi/4$) and the other a larger angle ($\theta_{23}$ > $\pi/4$). This ambiguity, known as the octant ambiguity of $\theta_{23}$, presents a significant challenge in neutrino research and is a primary objective of future long-baseline experiments. In this study, for the first time, we explore how non-isotropic Lorentz violation affects measurements of mixing angle $\theta_{23}$, with a particular emphasis on sidereal effects in the Deep Underground Neutrino Experiment. Our findings reveal that ability of DUNE to resolve the octant ambiguity is significantly compromised in the presence of the $c^{xy}_{e \tau}$ parameter. Furthermore, we demonstrate that LIV exacerbates the degeneracy between the Dirac CP-phase $\delta_{cp}$ and $\theta_{23}$.

hep-ph↗

arXiv2024

Investigating Lorentz Invariance Violation Effects on CP Violation and Mass Hierarchy sensitivity at DUNE

One of the current goals of neutrino experiments is to precisely determine standard unknown oscillation parameters such as the leptonic CP phase and mass hierarchy. Lorentz invariance violation represents a potential physics factor that could influence the experiment's ability to achieve these precise determinations. This study investigates the influence of Lorentz invariance violation (LIV) on oscillation dynamics, particularly through non-isotropic CPT-violating ($a^{X}_{e\mu}$, $a^{X}_{e\tau}$, $a^{X}_{\mu\tau}$) and CPT-conserving ($c^{XY}_{e\mu}$, $c^{XY}_{e \tau}$, $c^{XY}_{\mu \tau}$) parameters within the Deep Underground Neutrino Experiment (DUNE). We analyze the impact of these parameters on the mass hierarchy (MH) and Dirac CP phase sensitivity measurements. Our findings indicate that while MH sensitivity remains relatively unaffected, only the presence of $c^{XY}_{\mu \tau}$ significantly deteriorates MH sensitivity, albeit remaining above the $5 \sigma$ threshold. Additionally, we observe a substantial compromise in CP sensitivity due to the $c^{XY}_{e \mu}$ and $c^{XY}_{e \tau}$ parameters.

hep-ph↗

arXiv2024

Design and Implementation of Energy-Efficient Wireless Tire Sensing System with Delay Analysis for Intelligent Vehicles

The growing prevalence of Internet of Things (IoT) technologies has led to a rise in the popularity of intelligent vehicles that incorporate a range of sensors to monitor various aspects, such as driving speed, fuel usage, distance proximity and tire anomalies. Nowadays, real-time tire sensing systems play important roles for intelligent vehicles in increasing mileage, reducing fuel consumption, improving driving safety, and reducing the potential for traffic accidents. However, the current tire sensing system drains a significant vehicle' energy and lacks effective collection of sensing data, which may not guarantee the immediacy of driving safety. Thus, this paper designs an energy-efficient wireless tire sensing system (WTSS), which leverages energy-saving techniques to significantly reduce power consumption while ensuring data retrieval delays during real-time monitoring. Additionally, we mathematically analyze the worst-case transmission delay of the system to ensure the immediacy based on the collision probabilities of sensor transmissions. This system has been implemented and verified by the simulation and field trial experiments. These results show that the proposed scheme provides enhanced performance in energy efficiency and accurately identifies the worst transmission delay.

cs.ET↗

arXiv2023

Search for Lorentz-violation through sidereal effect at NO{\nu}A Experiment

Long-baseline neutrino oscillation experiments offer a unique laboratory to test the fundamental Lorentz symmetry, which is heart of both the standard model of particle and general relativity theory. Deviations from the standard neutrino oscillation or the sidereal modulation in neutrino events will smoking-gun experimental signature of Lorentz and CPT violation. In this study, we investigate the impact of the sidereal effect on standard neutrino oscillation measurements within the context of the NO{\nu}A experiment. Additionally, we assess the sensitivity of the NO{\nu}A experiment to detect Lorentz-violating interactions, taking into account the sidereal effect. Furthermore, we highlight potential of the NO{\nu}A experiment to set the new constraints on anisotropic Lorentz-violating parameters.

hep-ph↗

arXiv2023

RC-BEVFusion: A Plug-In Module for Radar-Camera Bird's Eye View Feature Fusion

Radars and cameras belong to the most frequently used sensors for advanced driver assistance systems and automated driving research. However, there has been surprisingly little research on radar-camera fusion with neural networks. One of the reasons is a lack of large-scale automotive datasets with radar and unmasked camera data, with the exception of the nuScenes dataset. Another reason is the difficulty of effectively fusing the sparse radar point cloud on the bird's eye view (BEV) plane with the dense images on the perspective plane. The recent trend of camera-based 3D object detection using BEV features has enabled a new type of fusion, which is better suited for radars. In this work, we present RC-BEVFusion, a modular radar-camera fusion network on the BEV plane. We propose BEVFeatureNet, a novel radar encoder branch, and show that it can be incorporated into several state-of-the-art camera-based architectures. We show significant performance gains of up to 28% increase in the nuScenes detection score, which is an important step in radar-camera fusion research. Without tuning our model for the nuScenes benchmark, we achieve the best result among all published methods in the radar-camera fusion category.

cs.CV↗

arXiv2021

Effect of entanglement and crosslinking on the hyperelastic behavior of SBR rubber: A multiscale DPD simulation study

In this study we have investigated into the entanglement effect and crosslinking effect in long-chain SBR rubber polymer system to model the mechanical hyperelastic behavior. The discussed methodologies are developed by a dissipative particle dynamics (DPD) based multiscale modeling method. The DPD interaction parameters are in turn obtained by all atomistic molecular dynamics and subsequently utilized by the DPD simulations. In the DPD simulation boxes, 200 long polymer chains and 1600 vulcanizing sulphur beads were packed. Vulcanization was achieved by random cross-linking among chains and entanglement was detected by utilizing the M-coil estimator. The mechanical effect of entanglement was modelled by extended tube model and the crosslinking effect was modelled by the Arruda-Boyce model. The modelled stress-strain curve is compared with the experimentally obtained curve for the qualitative and quantitative aspects of mechanical properties. It was determined that for strains up to 150%, the contribution of entanglement to the measured mechanical properties can be even up to 50%.

cond-mat.mtrl-sci↗

arXiv2021

Temperature dependent phase stability of Mo-Nb-Ta-W refractory high-entropy alloys

In this study Mo-Nb-Ta-W refractory high-entropy alloys (R-HEAs) have been studied for their phase stability for a wide temperature range (100 K to 2000 K). The equilibrium thermodynamic phases are determined by the changes in enthalpy and entropy. The enthalpy changes at any temperature are simulated by embedded atom method (EAM) potential based hybrid Monte Carlo molecular dynamics (MC/MD) simulations. Configurational entropy was calculated by quasichemical method. The EAM potentials were all calculated based on the physical input parameters of elements like atomic volume, cohesive energy, elastic constants etc. It was found that the MC/MD evolved equilibrium structures and the degree of local chemical short-range order/clustering (SRO/SRC) largely depend on the ordering enthalpies for various temperatures. The ordering transition temperature is close to 700 K, which is also substantiated by the experimental powder X-ray diffractions done in synchrotron beam. Large increase of degree of next-neighbor B2-type ordering of Mo-Ta pairs and decrease of that Nb-W pairs were observed. This was also expected from the trends of density functional theory (DFT) based ab-initio simulations done in the literature and discussed in this work. The simulated diffraction pattern and changes in scattering intensity due to the chemical SRO and SRC was investigated. The diffraction trends were explained with the help of the features of the evolved nanostructure morphology. The developed methodologies may be helpful in prior prediction of long-term phase stability in multi-element alloys by reducing the number of costly experiments.

cond-mat.mtrl-sci↗

arXiv2020

Computational property predictions of Ta-Nb-Hf-Zr high-entropy alloys

Refractory high entropy alloys (R-HEAs) are having properties and uses as high strength and high hardness materials for ambient and high temperature, aerospace and nuclear radiation tolerance applications, orthopedic applications etc. The mechanical properties like yield strength and ductility of TaNbHfZr R-HEA depend on the local nanostructure and chemical ordering. In this study we have computationally obtained various properties of the TaNbHfZr alloy like the role of configurational entropy in the thermodynamic property, rate of evolution of nanostructure morphology in thermally annealed systems, dislocation simulation based quantitative prediction of yield strength, nature of dislocation movement through short range clustering (SRC) and qualitative prediction of ductile to brittle transition behavior. The simulation starts with hybrid Monte Carlo/ Molecular Dynamics (MC/MD) based nanostructure evolution of an initial random solid solution alloy structure with BCC lattice structure created with principal axes along [1 1 1], [-1 1 0] and [-1 -1 2] directions suitable for simulation of 1/2[1 1 1] edge dislocations. Thermodynamic properties are calculated from the change in enthalpy and the configurational entropy by next-neighbor bond counting statistics. The MC/MD evolved structures mimic the annealing treatment at 1800{\deg}C and the output structures are replicated in periodic directions to make larger 384000 atom structures used for dislocation simulations. Edge dislocations were utilized to obtain and explain for the extra strengthening observed because of the formations of SRCs. Lastly the MC/MD evolved structures containing dislocations are subjected to a high shear stress beyond CRSS to investigate the stability of the dislocations and the lattice structures to explain the experimentally observed transition from ductile to brittle behavior for the TaNbHfZr R-HEA.

cond-mat.mtrl-sci↗