Searcharxiv⌕ Search

arXiv subjects

Nikil Dutt

Publications and source records attributed to Nikil Dutt.

At least 19 recordsLinked to original sources
arXiv2026

FloodReasonBench: Benchmarking VLM Reasoning Segmentation for Embodied Flood Response at the Edge

Reasoning segmentation enables vision-language models (VLMs) to translate mission-relevant language requests into pixel-level visual grounding, offering a natural perception interface for embodied agents. However, existing benchmarks largely focus on generic visual scenes and overlook the domain and resource constraints encountered in flood-response platforms. We present FloodReasonBench, a benchmark for VLM reasoning segmentation for embodied flood response at the edge. At its core, FloodReasonBench introduces FloodResponseSeg, a flood-specific reasoning-segmentation dataset constructed from real-world scenes and response-relevant targets. Beyond task accuracy, the benchmark characterizes reasoning-segmentation pipelines under lightweight visual encoding, hierarchical split inference, and compressed intermediate representations. We observe strong partition-dependent accuracy variation in the generic pre-adaptation setting, while the flood-adapted target-workload design space exhibits a substantially more compact accuracy range across partitions. Evaluation on an NVIDIA Jetson AGX Xavier further exposes the tradeoffs among reasoning-segmentation accuracy, edge-side latency, energy, and communication footprint, enabling quality-constrained selection of edge operating points. Together, these results provide a task- and system-level characterization of reasoning segmentation for resource-constrained embodied flood response at the edge.

cs.DC↗
arXiv2026

MemoGuard: An Adaptive Runtime for Guarding Against Memory Traps in Communication-Limited Robot Navigation

Communication-limited robots in mission-critical scenarios such as disaster inspection and search-and-rescue must make reliable onboard decisions without access to remote operators or high-capacity reasoning services. Episodic memory reuse is an attractive low-cost fallback, but retrieval similarity does not guarantee execution validity, i.e., a retrieved action may match the current context yet be unsafe due to changed topology, insufficient battery margin, or unreliable prior outcomes. We call such high-similarity but execution-invalid episodes memory traps. This creates a safety-efficiency design space where similarity only reuse minimizes fallback cost but can be unsafe, while always invoking local reasoning improves safety at high computational and energy cost. This paper presents MemoGuard, a lightweight adaptive runtime that validates episodic memories against topology, resource, and outcome contracts before reuse, invoking fallback only when validation fails. In a graph-based corridor-inspection simulator, MemoGuard reduces battery safety violations by 76.6% over similarity-only top-1 reuse while reducing fallback calls by 21.4% over always reasoning. On an NVIDIA Jetson AGX Xavier with local llama3.2:3b fallback reasoning, this corresponds to 3.67 s and 36.97 J of avoided fallback-reasoning overhead per trial. We open-source MemoGuard at https://github.com/hetheiin/memoguard.

cs.RO↗
arXiv2026

Can the Environment Speak for Itself? $T^{2}$-GRPO: A Turn-Trajectory Group Relative Policy Optimization for Caregiver Agents

Optimizing large language models (LLMs) for long-horizon caregiver agents requires balancing delayed task objectives with immediate environment dynamics, such as patient distress and resistance. In dementia care, this balance is especially difficult: trajectory level rewards are too sparse for turn level credit assignment, while external LLM-based evaluators are costly and can misread fragmented or indirect patient responses. To address this issue, we propose \textbf{T}urn-\textbf{T}rajectory \textbf{G}roup \textbf{R}elative \textbf{P}olicy \textbf{O}ptimization (\textbf{T$^{2}$-GRPO}), a framework that decouples caregiver RL into two normalized reward horizons and enforces safety through a binary hard veto. $T^2$-GRPO derives dense turn-level rewards directly from environment state transitions, measuring changes in patient distress and resistance from a frozen dementia patient simulator. These environment-grounded rewards are combined with trajectory-level evaluations through independent centered-rank normalization, which preserves heterogeneous reward signals and mitigates reward collapse. Extensive experiments on dementia caregivers show that T $^{2}$-GRPO outperforms competitive baselines, indicating a substantial improvement for emotionally sensitive caregiver scenarios that effectively handles immediate patient feedback, long-term care outcomes, and safety constraints.

cs.AI↗
arXiv2026

Orion: Enabling Self-adaptive Memory Management for On-device Online Continual Learning

Online continual learning (OCL) enables real-time adaptation to new data, making it crucial for dynamic robotic applications. However, its practical deployment is hindered by memory constraints in resource-limited systems, which affect key trade-offs in training latency, plasticity, and stability. Unlike offline parameter tuning, which cannot account for the dynamic shift in memory pressure and workload complexity as OCL progresses, an online and self-adaptive approach is essential for robust on-device deployment. This paper proposes Orion, a holistic framework designed to co-optimize training latency, plasticity, and stability of state-of-the-art OCL models under strict memory constraints, enabling feasible on-device deployment. At its core, Orion leverages URGE, a unified runtime indicator grounded in the ``Buckets effect'' principle that system performance is bounded by its scarcest resource, to dynamically reallocate memory across OCL components by jointly coordinating batch processing, replay buffers, and optimization strategies at both the OS and application level. Furthermore, Orion introduces system-level data prefetching techniques to maximize efficiency. A system prototype of Orion has been implemented using the widely adopted \texttt{Avalanche-lib} and thoroughly evaluated across a diverse range of OCL algorithms, benchmarks, and hardware platforms commonly used in autonomous robotic applications. To further demonstrate its practical utility, Orion is integrated into a realistic autonomous navigational robot powered by OCL. The results show that Orion achieves significant training speedups while maintaining balanced performance and effectively adapting to various scenarios, all with minimal runtime, memory, and energy overhead, making Orion a practical solution for on-device continual learning.

eess.SY↗
arXiv2026

GeoReFormer: Geometry-Aware Refinement for Lane Segment Detection and Topology Reasoning

Accurate 3D lane segment detection and topology reasoning are critical for structured online map construction in autonomous driving. Recent transformer-based approaches formulate this task as query-based set prediction, yet largely inherit decoder designs originally developed for compact object detection. However, lane segments are continuous polylines embedded in directed graphs, and generic query initialization and unconstrained refinement do not explicitly encode this geometric and relational structure. We propose GeoReFormer (Geometry-aware Refinement Transformer), a unified query-based architecture that embeds geometry- and topology-aware inductive biases directly within the transformer decoder. GeoReFormer introduces data-driven geometric priors for structured query initialization, bounded coordinate-space refinement for stable polyline deformation, and per-query gated topology propagation to selectively integrate relational context. On the OpenLane-V2 benchmark, GeoReFormer achieves state-of-the-art performance with 34.5% mAP while improving topology consistency over strong transformer baselines, demonstrating the utility of explicit geometric and relational structure encoding.

cs.CV↗
arXiv2026

DementiaBank-Emotion: A Multi-Rater Emotion Annotation Corpus for Alzheimer's Disease Speech (Version 1.0)

We present DementiaBank-Emotion, the first multi-rater emotion annotation corpus for Alzheimer's disease (AD) speech. Annotating 1,492 utterances from 108 speakers for Ekman's six basic emotions and neutral, we find that AD patients express significantly more non-neutral emotions (16.9%) than healthy controls (5.7%; p < .001). Exploratory acoustic analysis suggests a possible dissociation: control speakers showed substantial F0 modulation for sadness (Delta = -3.45 semitones from baseline), whereas AD speakers showed minimal change (Delta = +0.11 semitones; interaction p = .023), though this finding is based on limited samples (sadness: n=5 control, n=15 AD) and requires replication. Within AD speech, loudness differentiates emotion categories, indicating partially preserved emotion-prosody mappings. We release the corpus, annotation guidelines, and calibration workshop materials to support research on emotion recognition in clinical populations.

cs.CL↗
arXiv2026

MedSpeak: A Knowledge Graph-Aided ASR Error Correction Framework for Spoken Medical QA

Spoken question-answering (SQA) systems relying on automatic speech recognition (ASR) often struggle with accurately recognizing medical terminology. To this end, we propose MedSpeak, a novel knowledge graph-aided ASR error correction framework that refines noisy transcripts and improves downstream answer prediction by leveraging both semantic relationships and phonetic information encoded in a medical knowledge graph, together with the reasoning power of LLMs. Comprehensive experimental results on benchmarks demonstrate that MedSpeak significantly improves the accuracy of medical term recognition and overall medical SQA performance, establishing MedSpeak as a state-of-the-art solution for medical SQA. The code is available at https://github.com/RainieLLM/MedSpeak.

cs.CL↗
arXiv2026

Do Implicit Personalization and Explicit Styles Conflict? PsPLUG: A Lightweight Plug-in for Balancing Personalization and Style in Customized LLMs

Personalized large language models are often expected to follow explicit style instructions, yet we find that such instructions can undermine the user-specific characteristics that personalization methods aim to preserve. We call this failure mode personalization collapse: explicit style control can conflict with implicit user preferences. To address this challenge, we propose PsPLUG, a lightweight plug-in that learns a user-specific residual after accounting for the requested style. PsPLUG also allows us to tune personalization strength at inference time. Our experiments show that explicit style instructions can diminish personalization in existing methods, whereas PsPLUG better preserves user preferences while providing precise control over the balance between personalization and style adherence.

cs.AI↗
arXiv2026

DemMA: Dementia Multi-Turn Dialogue Agent with Expert-Guided Reasoning and Action Simulation

Simulating dementia patients with large language models (LLMs) is challenging due to the need to jointly model cognitive impairment, emotional dynamics, and nonverbal behaviors over long conversations. We present DemMA, an expert-guided dementia dialogue agent for high-fidelity multi-turn patient simulation. DemMA constructs clinically grounded dementia personas by integrating pathology information, personality traits, and subtype-specific memory-status personas informed by clinical experts. To move beyond text-only simulation, DemMA explicitly models nonverbal behaviors, including motion, facial expressions, and vocal cues. We further introduce a Chain-of-Thought distillation framework that trains a single LLM to jointly generate reasoning traces, patient utterances, and aligned behavioral actions within one forward pass, enabling efficient deployment without multi-agent inference. Extensive evaluations with experts, medical students, and LLM judges demonstrate that DemMA significantly outperforms strong baselines across multiple metrics.

cs.MA↗
arXiv2026

CARD: Cluster-level Adaptation with Reward-guided Decoding for Personalized Text Generation

Adapting large language models to individual users remains challenging due to the tension between fine-grained personalization and scalable deployment. We present CARD, a hierarchical framework that achieves effective personalization through progressive refinement. CARD first clusters users according to shared stylistic patterns and learns cluster-specific LoRA adapters, enabling robust generalization and strong low-resource performance. To capture individual differences within each cluster, we propose an implicit preference learning mechanism that contrasts user-authored text with cluster-level generations, allowing the model to infer user-specific style preferences without manual annotation. At inference time, CARD injects personalization exclusively at decoding via lightweight user preference vectors and low-rank logit corrections, while keeping the base model frozen. Experiments on the LaMP and LongLaMP benchmarks show that CARD achieves competitive or superior generation quality compared to state-of-the-art baselines, while significantly improving efficiency and scalability for practical personalized text generation.

cs.AI↗
arXiv2025

AVERY: Intent-Driven Adaptive VLM Split Computing via Embodied Self-Awareness for Efficient Disaster Response Systems

Unmanned Aerial Vehicles (UAVs) in disaster response require complex, queryable intelligence that onboard CNNs cannot provide. While Vision-Language Models (VLMs) offer this semantic reasoning, their high resource demands make on-device deployment infeasible, and naive cloud offloading fails under the low-bandwidth, unstable networks endemic to disaster zones. We present AVERY, an intent-driven adaptive split computing framework for efficient VLM deployment on resource-constrained platforms. AVERY is motivated by the observation that operator intent must be treated as a first-class system objective, since missions such as broad situational monitoring and precise, spatially grounded investigation require different semantic products, latency targets, and resource allocations. To reflect this, AVERY advances split computing beyond traditional depth-wise partitioning through a functional, cognitive-inspired dual-stream split: a high-frequency, low-resolution Context stream for real-time awareness, and a low-frequency, high-fidelity Insight stream for deep analysis. This design enables a hierarchical split strategy: computation is first separated by function, then partitioned depth-wise across edge and cloud when the Insight stream is required. A lightweight, self-aware onboard controller monitors network conditions and operator intent to select from pre-trained compression models, navigating the accuracy-throughput trade-off at runtime. Evaluated using LISA-7B in an edge-cloud setting under fluctuating network conditions, AVERY achieves 11.2% higher accuracy than raw image compression, 93.98% lower energy consumption than full-edge execution, and average accuracy within 0.75% of the static High-Accuracy baseline during dynamic adaptation. Overall, AVERY enhances mission efficiency and enables real-time, queryable intelligence in dynamic disaster environments.

cs.DC↗
arXiv2025

T-SAR: A Full-Stack Co-design for CPU-Only Ternary LLM Inference via In-Place SIMD ALU Reorganization

Recent advances in LLMs have outpaced the computational and memory capacities of edge platforms that primarily employ CPUs, thereby challenging efficient and scalable deployment. While ternary quantization enables significant resource savings, existing CPU solutions rely heavily on memory-based lookup tables (LUTs) which limit scalability, and FPGA or GPU accelerators remain impractical for edge use. This paper presents T-SAR, the first framework to achieve scalable ternary LLM inference on CPUs by repurposing the SIMD register file for dynamic, in-register LUT generation with minimal hardware modifications. T-SAR eliminates memory bottlenecks and maximizes data-level parallelism, delivering 5.6-24.5x and 1.1-86.2x improvements in GEMM latency and GEMV throughput, respectively, with only 3.2% power and 1.4% area overheads in SIMD units. T-SAR achieves up to 2.5-4.9x the energy efficiency of an NVIDIA Jetson AGX Orin, establishing a practical approach for efficient LLM inference on edge platforms.

cs.AR↗
arXiv2025

FlexiCache: Leveraging Temporal Stability of Attention Heads for Efficient KV Cache Management

Large Language Model (LLM) serving is increasingly constrained by the growing size of the key-value (KV) cache, which scales with both context length and generation length. Prior work shows that attention is dominated by a small subset of critical tokens, yet existing systems struggle to exploit this efficiently without degrading accuracy, especially in long generation. We make a key observation: the temporal stability of these critical tokens varies significantly across KV heads: some heads consistently focus on the same tokens, while others shift frequently. Building on this insight, we introduce FlexiCache, a hierarchical KV-cache management system that leverages the temporal stability of KV heads to reduce GPU memory usage and computation overhead, while preserving model accuracy. FlexiCache classifies KV heads as stable or unstable: it retains all KV-cache pages from unstable heads in GPU memory, whereas for stable heads, it keeps only the top-K pages on the GPU and offloads the rest to host memory. By exploiting temporal stability, FlexiCache performs periodic reranking for stable heads to fetch newly promoted top pages. Implemented atop vLLM, FlexiCache reduces GPU memory footprint for long-context requests by up to 70%, improves offline serving throughput by 1.38-1.55x, and lowers online token latency by 1.6-2.1x, all while maintaining accuracy in long-context, long-generation scenarios.

cs.LG↗
arXiv2025

HYPERDOA: Robust and Efficient DoA Estimation using Hyperdimensional Computing

Direction of Arrival (DoA) estimation techniques face a critical trade-off, as classical methods often lack accuracy in challenging, low signal-to-noise ratio (SNR) conditions, while modern deep learning approaches are too energy-intensive and opaque for resource-constrained, safety-critical systems. We introduce HYPERDOA, a novel estimator leveraging Hyperdimensional Computing (HDC). The framework introduces two distinct feature extraction strategies -- Mean Spatial-Lag Autocorrelation and Spatial Smoothing -- for its HDC pipeline, and then reframes DoA estimation as a pattern recognition problem. This approach leverages HDC's inherent robustness to noise and its transparent algebraic operations to bypass the expensive matrix decompositions and "black-box" nature of classical and deep learning methods, respectively. Our evaluation demonstrates that HYPERDOA achieves ~35.39% higher accuracy than state-of-the-art methods in low-SNR, coherent-source scenarios. Crucially, it also consumes ~93% less energy than competing neural baselines on an embedded NVIDIA Jetson Xavier NX platform. This dual advantage in accuracy and efficiency establishes HYPERDOA as a robust and viable solution for mission-critical applications on edge devices.

eess.SP↗
arXiv2025

ACCESS-AV: Adaptive Communication-Computation Codesign for Sustainable Autonomous Vehicle Localization in Smart Factories

Autonomous Delivery Vehicles (ADVs) are increasingly used for transporting goods in 5G network-enabled smart factories, with the compute-intensive localization module presenting a significant opportunity for optimization. We propose ACCESS-AV, an energy-efficient Vehicle-to-Infrastructure (V2I) localization framework that leverages existing 5G infrastructure in smart factory environments. By opportunistically accessing the periodically broadcast 5G Synchronization Signal Blocks (SSBs) for localization, ACCESS-AV obviates the need for dedicated Roadside Units (RSUs) or additional onboard sensors to achieve energy efficiency as well as cost reduction. We implement an Angle-of-Arrival (AoA)-based estimation method using the Multiple Signal Classification (MUSIC) algorithm, optimized for resource-constrained ADV platforms through an adaptive communication-computation strategy that dynamically balances energy consumption with localization accuracy based on environmental conditions such as Signal-to-Noise Ratio (SNR) and vehicle velocity. Experimental results demonstrate that ACCESS-AV achieves an average energy reduction of 43.09% compared to non-adaptive systems employing AoA algorithms such as vanilla MUSIC, ESPRIT, and Root-MUSIC. It maintains sub-30 cm localization accuracy while also delivering substantial reductions in infrastructure and operational costs, establishing its viability for sustainable smart factory environments.

eess.SY↗
arXiv2025

DEMENTIA-PLAN: An Agent-Based Framework for Multi-Knowledge Graph Retrieval-Augmented Generation in Dementia Care

Mild-stage dementia patients primarily experience two critical symptoms: severe memory loss and emotional instability. To address these challenges, we propose DEMENTIA-PLAN, an innovative retrieval-augmented generation framework that leverages large language models to enhance conversational support. Our model employs a multiple knowledge graph architecture, integrating various dimensional knowledge representations including daily routine graphs and life memory graphs. Through this multi-graph architecture, DEMENTIA-PLAN comprehensively addresses both immediate care needs and facilitates deeper emotional resonance through personal memories, helping stabilize patient mood while providing reliable memory support. Our notable innovation is the self-reflection planning agent, which systematically coordinates knowledge retrieval and semantic integration across multiple knowledge graphs, while scoring retrieved content from daily routine and life memory graphs to dynamically adjust their retrieval weights for optimized response generation. DEMENTIA-PLAN represents a significant advancement in the clinical application of large language models for dementia care, bridging the gap between AI tools and caregivers interventions.

cs.AI↗
arXiv2024

Loneliness Forecasting Using Multi-modal Wearable and Mobile Sensing in Everyday Settings

The adverse effects of loneliness on both physical and mental well-being are profound. Although previous research has utilized mobile sensing techniques to detect mental health issues, few studies have utilized state-of-the-art wearable devices to forecast loneliness and estimate the physiological manifestations of loneliness and its predictive nature. The primary objective of this study is to examine the feasibility of forecasting loneliness by employing wearable devices, such as smart rings and watches, to monitor early physiological indicators of loneliness. Furthermore, smartphones are employed to capture initial behavioral signs of loneliness. To accomplish this, we employed personalized machine learning techniques, leveraging a comprehensive dataset comprising physiological and behavioral information obtained during our study involving the monitoring of college students. Through the development of personalized models, we achieved a notable accuracy of 0.82 and an F-1 score of 0.82 in forecasting loneliness levels seven days in advance. Additionally, the application of Shapley values facilitated model explainability. The wealth of data provided by this study, coupled with the forecasting methodology employed, possesses the potential to augment interventions and facilitate the early identification of loneliness within populations at risk.

eess.SP↗
arXiv2024

Enhancing Performance and User Engagement in Everyday Stress Monitoring: A Context-Aware Active Reinforcement Learning Approach

In today's fast-paced world, accurately monitoring stress levels is crucial. Sensor-based stress monitoring systems often need large datasets for training effective models. However, individual-specific models are necessary for personalized and interactive scenarios. Traditional methods like Ecological Momentary Assessments (EMAs) assess stress but struggle with efficient data collection without burdening users. The challenge is to timely send EMAs, especially during stress, balancing monitoring efficiency and user convenience. This paper introduces a novel context-aware active reinforcement learning (RL) algorithm for enhanced stress detection using Photoplethysmography (PPG) data from smartwatches and contextual data from smartphones. Our approach dynamically selects optimal times for deploying EMAs, utilizing the user's immediate context to maximize label accuracy and minimize intrusiveness. Initially, the study was executed in an offline environment to refine the label collection process, aiming to increase accuracy while reducing user burden. Later, we integrated a real-time label collection mechanism, transitioning to an online methodology. This shift resulted in an 11% improvement in stress detection efficiency. Incorporating contextual data improved model accuracy by 4%. Personalization studies indicated a 10% enhancement in AUC-ROC scores, demonstrating better stress level differentiation. This research marks a significant move towards personalized, context-driven real-time stress monitoring methods.

cs.LG↗