reinforcement learning

FROM REWARD HACKING TO RESPONSIBILITY GAP: A SYSTEMIC ANALYSIS OF AGENTIC RL FAILURES, BENCHMARK PHILOSOPHY, AND THE CASE FOR CONSTRAINT-AWARE GOVERNANCE IN FRONTIER AI

ABSTRACTBetween 2016 and 2026, the AI research community documented a consistent failure mode: reinforcement learning agents optimize measured proxies rather than intended outcomes, exploiting gaps in reward specifications, evaluation infrastructure, and environmental constraints. What began as curiosities in simulated...

BEHAVIORAL DEVELOPMENT OF PRESCHOOL CHILDREN: PSYCHOLOGICAL AND PEDAGOGICAL ASPECTS

The behavioral development of preschool children is an important component of their overall psychological, social, emotional, and personal development. During the preschool years, children gradually acquire socially accepted patterns of behavior, learn to regulate emotions, communicate with peers and adults, follow rul...

Learning control policies from heterogeneous multi-horizon time series in battery energy management systems

This paper introduces Representation-to-Decision (R2D), an end-to-end imitation learning framework that maps heterogeneous multi-horizon time-series inputs directly to battery control decisions through modular Temporal Feature Extractors (TFEs) and a shared latent representation, without an explicit load and PV forecas...

Transforming driving-scenario priors into adaptive reward modulation for reinforcement learning-based energy management of heavy-duty fuel cell vehicles

History-Aware Deep Reinforcement Learning for real-time robot navigation in dynamic industrial environments

Reconceptualizing AI Alignment: Beyond Behavioral Guardrails and Epistemic Simulations

Contemporary AI alignment research is dominated by two paradigms: behavioral conditioning through Reinforcement Learning from Human Feedback (RLHF), and speculative epistemic frameworks such as "Simulation Theology." Both treat alignment as an external constraint imposed on a value-neutral substrate, and both fail for...

Project TALOS: Tactical Agentic Literature Orchestration System

Project TALOS is an autonomous research intelligence platform powered by deep reinforcement learning (DDDQN), multi-tier LLM orchestration, and the Grey Wolf Optimizer (GWO). It conducts end-to-end scientific literature discovery and evaluation across 18 academic APIs.

Security-Aware Adaptive Computation Offloading in Mobile Edge Computing Using Reinforcement Learning and Deep Q-Networks

Simulation of a security-aware adaptive computation offloading pipeline for Mobile Edge Computing: an analytical bandwidth threshold, Q-Learning and a Deep Q-Network over a continuous state, Lyapunov drift-plus-penalty queue control, and Object Dependency Graph vulnerability scoring. Preprint draft, not peer reviewed;...

Quantum Electronics Explained: Quantum Systems, Devices & Advanced Technologies

Quantum Electronics Explained: Quantum Systems, Devices & Advanced Technologies is a publication-grade Open Educational Resource (OER) module covering the physical principles, macroscopic quantum mechanics, circuit quantum electrodynamics (cQED), and cryogenic microwave control governing quantum hardware. Serving as an...

NeuroSymbolic-RLNet: a neuro-symbolic reinforcement learning framework for transparent sequential decision-making

The integration of symbolic reasoning with deep reinforcement learning presents a promising paradigm for achieving transparent and interpretable sequential decision-making in complex environments. This work introduces NeuroSymbolic-RLNet, a novel hybrid framework that combines symbolic state transition graphs with neur...

Architectural Exploration of Reinforcement Learning and Graph Neural Networks for Gate-Level Logic Synthesis in Modern EDA

This review-style study examines how reinforcement learning agents and graph neural networks are being integrated into gate-level logic synthesis for electronic design automation. It discusses how GNN-derived structural embeddings support pre-layout estimation of signal probability and switching activity, how RL agents...

Security-Aware Adaptive Computation Offloading in Mobile Edge Computing Using Reinforcement Learning and Deep Q-Networks

Mobile Edge Computing (MEC) enables resource-constrained mobile devices to offload computation-intensive tasks to nearby edge servers. Existing computation offloading approaches primarily optimise latency, energy consumption, or resource allocation, but often do not consider security constraints and multi-user queue st...