Large language models (LLMs) can struggle with time-series question answering (TS-QA), especially when numerical signals are serialized as text and require explicit computation. Tool-augmented approaches improve performance, but existing systems often intertwine language reasoning, computation, and perception, making it difficult to determine which components drive the gains. We present T-SMART, a neurosymbolic framework that separates these roles: a frozen LLM interprets questions and selects operations, deterministic tools perform numerical computation, and structured perception is invoked only when needed. Controlled paired ablations show that deterministic computation provides the dominant benefit, improving accuracy by 31.7 percentage points over direct LLM reasoning on serialized time series, while language understanding and perception offer smaller complementary gains. These results indicate that tool-augmented TS-QA benefits primarily from reliable numerical execution rather than additional language-model reasoning and provide a controlled framework for analyzing component contributions in neurosymbolic time-series systems.
I. Delgado, Himansi Gupta, B. Khatri et al.· 0 citations
Minimizing computational overhead in healthcare time-series classification remains a critical challenge, particularly for deep learning models operating on high-dimensional sequential data under resource and latency constraints. This challenge is further exacerbated by adversarial attacks, which introduce evolving threats and necessitate robust yet efficient mechanisms for maintaining reliable model performance—a requirement of paramount importance in healthcare, where wearable and smart devices have brought continuous monitoring outside clinical settings and model failures can directly compromise patient safety. In this paper, we propose ReLATE+, a unified framework for adversarially robust and computationally efficient time-series classification. ReLATE+ integrates three key capabilities: (i) detection and classification of adversarial inputs, (ii) dataset-level similarity analysis, and (iii) adaptive model selection. Upon receiving new data, the framework first determines whether the input is adversarial and identifies the attack type. It then leverages this information to retrieve a similar dataset from a repository and identify the corresponding high-performing models and trains only a small set of selected candidates instead of exhaustively retraining all models. This approach ensures strong performance while reducing the need for retraining, and it generalizes well across different domains with varying data distributions and feature spaces. Experiments show that ReLATE+ reduces computational overhead by an average of 77.68%, enhancing adversarial resilience and streamlining robust model selection, all without sacrificing performance, within 2.02% of Oracle.
C. I. Kocal, Onat Gungor, T. Rosing et al.· Electronics· 0 citations
Autonomous driving systems must operate under partial observability, where safety-critical objects may be occluded or visible only to neighboring connected vehicles. Vehicle-to-vehicle cooperation can reduce this uncertainty, but existing cooperative driving methods often compress multi-agent evidence into latent features or hidden multimodal states. As a result, they obscure which agent observed each object, whether the object is visible to the ego vehicle, and how conflicting evidence affects downstream decisions. We propose G-MARK, a grounded multi-agent reasoning framework that converts cooperative object-centric observations into explicit provenance-aware knowledge graphs (KGs). The resulting KGs preserve object hypotheses together with their source attribution, ego-versus-partner visibility, uncertainty, conflicts, spatial relations, and planning-relevant context. G-MARK then derives a shared feature representation from these KGs, enabling lightweight task heads to support object reasoning, motion prediction, control selection, and trajectory forecasting. Compared with the state-of-the-art baseline, GMARK improves occlusion reasoning accuracy by 42.2%, reduces control-selection error by 13.1%, and achieves comparable trajectory-planning accuracy with a 25.6x smaller structured communication payload. Our code is available at https://github.com/bhavyagupta98/g-mark.
