arXiv subjects

Sokipriala Jonah

Publications and source records attributed to Sokipriala Jonah.

5 recordsLinked to original sources

arXiv2026

Controlled Evaluation of Graph and Multimodal Augmentation in RAG for Document Question Answering

Graph and multimodal extensions to retrieval-augmented generation (RAG) are often evaluated end to end, making it difficult to isolate whether gains arise from retrieval, prompt-side context, visual access, generator capability, or benchmark construction. We present a stage- and evidence-controlled evaluation across five RAG configurations, four multimodal generators, and three document corpora. The same LLM-extracted knowledge graph is used either after retrieval as provenance-constrained triple injection (+KG) or during retrieval as entity-bridged passage expansion (+KGret). Prompt-side graph injection yields no consistent accuracy improvement and generally reduces faithfulness. In contrast, +KGret increases gold-evidence completeness from 0.22 to 0.46 on HotpotQA bridge questions and from 0.50 to 0.72 on SPIQA cross-paper questions, improving accuracy for every generator on both evidence-deficient sets while having little effect on retrieval-complete controls. For visual question answering, matched caption-answerable and verified pixel-only protocols show that apparent multimodal gains are sensitive to textual leakage. Programmatic checks reveal answer recoverability from captions, corpus text, and model responses generated without complete gold evidence. Accuracy on incomplete-evidence questions reaches 0.35--0.71 on widely disseminated corpora, compared with 0 on PubLayNet, indicating that raw accuracy can overstate retrieval-attributable performance. These results show that graph augmentation is most effective when it changes retrieval under evidence deficits, while multimodal evaluation requires explicit verification that answers are unavailable through text.

eess.IV↗

arXiv2026

LLMs for Agentic Home Energy Management

Home Energy Management Systems (HEMS) can reduce residential electricity costs, but many require users to express everyday preferences as technical constraints. This paper presents a tool-calling ReAct agent that converts natural-language requests into schedules for multiple household appliances using half-hourly Octopus Agile prices, weather forecasts, photovoltaic generation estimates, and household demand data. Five large language model backends are evaluated against a mixed-integer linear programming benchmark across dynamic tariff conditions, constraint conflicts, weather-aware scheduling, and a seven-day rolling deployment. Native function calling achieves high scheduling success and near-optimal cost on ordinary tariff days, whereas text-parsed actions reduce reliability. Constraint-conflict testing shows that low cost does not guarantee safe or feasible behaviour. Claude Sonnet 4.6 performs best in power-cap and infeasibility scenarios, while Qwen-3 achieves higher overall constraint compliance than GPT-4o-mini. The evaluation also identifies fabricated schedules, failed commitments, and reasoning-to-action failures in which models explain a deadline correctly but commit an invalid schedule. Weather-aware scheduling reduces cost and increases solar self-consumption under overcast conditions, but provides limited or adverse economic value under some dynamic-price regimes. Across the evaluated seven-day period, the agents capture 96.7-98.0% of the savings available between an off-peak timer and the MILP oracle and outperform the rule-based policies. The results support LLM-based HEMS orchestration, provided that every committed schedule is checked by an independent deterministic feasibility validator before actuation. Code and a live demonstration are available at https://github.com/sokistar24/ecohome-experiments and https://www.ecohomeagent.com/.

eess.SY↗

arXiv2026

Fair and Efficient Scheduling for Sensor Networks via Online Whittle Index Policy

Wake-Up Radio (WUR) enables resource-constrained, battery-powered sensor nodes to remain in a low-power deep sleep state while continuously listening for a Wake-Up Signal (WUS). Sensor nodes only wake and transmit data after receiving the WUS, significantly reducing energy consumption. However, polling nodes whose transmitted data provides little or no meaningful update to the remote monitor can still result in unnecessary energy usage and increased storage overhead. To address this issue, this paper uses the Age of Incorrect Information (AoII) metric to prioritise the polling of nodes that provide informative updates to the remote monitor. Determining the optimal set of nodes to poll based on AoII can be formulated as a Restless Multi-Armed Bandit (RMAB) problem, which traditionally requires prior knowledge of the monitored process transition dynamics. Since such dynamics are often unknown in practical deployments, we propose an online learning framework based on state estimation to derive Whittle Index AoII (WAoII) and Fair Whittle Index AoII (FWAoII) policies without assuming known transition probabilities. The proposed policies efficiently schedule node polling while adapting to unknown process behaviour. Experimental evaluation using both real-world and synthetic datasets demonstrates that the proposed online WAoII policy can reduce packet transmissions by up to 70\% compared to the widely used Round Robin (RR) polling strategy, while maintaining Root Mean Squared Error (RMSE) values within acceptable application error tolerances. These results demonstrate the effectiveness of WAoII and FWAoII as energy-efficient polling techniques for low-power WUR sensor networks.

eess.SP↗

arXiv2026

Empirical and Statistical Characterisation of 28 GHz mmWave Propagation in Office Environments

Millimeter wave (mmWave) technology at 28 GHz is vital for beyond-5G systems, but indoor deployment remains challenging due to limited statistical evidence on propagation. This study investigates path loss, material penetration, and coverage enhancement using TMYTEK-based measurements. Statistical tests and confidence interval analysis show that path loss aligns with free-space theory, with an exponent of n = 2.07 plus or minus 0.073 (p = 0.385), confirming the suitability of classical models. Material analysis reveals significant variation: desk dividers introduce 3.4 dB more attenuation than display boards (95 percent CI: 1.81 to 4.98 dB, p less than 0.01), contradicting thickness-based assumptions. Reflector optimisation yields a significant mean gain of 2.17 plus or minus 2.33 dB (p less than 0.05), enhancing coverage. The results provide new empirical benchmarks and practical design insights for reliable indoor mmWave deployment.

eess.SP↗

arXiv2026

Adaptive Scheduling: A Reinforcement Learning Whittle Index Approach for Wireless Sensor Networks

We propose a reinforcement learning based scheduling framework for Restless Multi-Armed Bandit (RMAB) problems, centred on a Whittle Index Q-Learning policy with Upper Confidence Bound (UCB) exploration, referred to as WIQL-UCB. Unlike existing approaches that rely on fixed or adaptive epsilon-greedy strategies and require careful hyperparameter tuning, the proposed method removes problem-specific tuning and is therefore more generalisable across diverse RMAB settings. We evaluate WIQL-UCB on standard RMAB benchmarks and on a practical sensor scheduling application based on the Age of Incorrect Information (AoII), using an edge-based state estimation scheme that requires no prior knowledge of system dynamics. Experimental results show that WIQL-UCB achieves near-optimal performance while significantly improving computational and memory efficiency. For a representative problem size of N = 15 and M = 3, the proposed method requires only around 600 bytes of memory, compared with several kilobytes for tabular Q-learning and hundreds of kilobytes to megabytes for deep reinforcement learning baselines. In addition, WIQL-UCB achieves sub-millisecond per-decision runtimes and is several times faster than deep reinforcement learning approaches, while maintaining competitive performance. Overall, these results demonstrate that WIQL-UCB consistently outperforms both non-Whittle-based and Whittle-index learning baselines across a wide range of RMAB settings.

eess.SY↗