2833 papers
This paper presents the first comprehensive, multi-dimensional survey of OpenClaw, a rapidly adopted yet security-vulnerable autonomous AI agent framework, analyzing its architecture, ecosystem, and governance to reveal the inherent structural tensions between accessibility, speed, and security in the emerging paradigm of autonomous AI agents.
This paper introduces the eligibility trace as a novel computational substrate that embeds the causal arrow of time directly into hardware, challenging the von Neumann architecture's assumption of time as a mere sequence number by demonstrating its unique necessity for emergent differentiation in spiking neural networks and its isomorphic causal structure in stochastic gradient descent.
This controlled empirical study demonstrates that framing Large Language Model responses for an adversarial fact-checker, rather than the direct asker, systematically reduces structural pathologies like sycophancy and fabrication by activating latent knowledge, an effect that is robust across models but imperceptible to human users, necessitating system-side implementation.
This paper proposes MCFN, a lightweight multi-path compound fusion network that integrates texture, edge, and frequency cues via cross-path attention and FiLM modulation to achieve state-of-the-art deepfake detection accuracy across multiple benchmarks with only 6.33 million parameters.
This paper introduces Analog Diffusion Models (ADMs), which leverage a fixed-point correspondence between analog hardware dynamics and implicit diffusion inference to enable fully analog, high-efficiency generative AI without requiring model redesign or compromising compatibility with standard training ecosystems.
The paper introduces OPCNet, a novel neural network model that identifies and fits highly oscillatory data by learning underlying linear ODE dynamics with perturbations, demonstrating superior performance in capturing oscillatory patterns compared to existing methods.
This paper addresses the "Insight-to-Action Gap" in intelligent systems by proposing a closed-loop framework that embeds governance and execution constraints directly into the decision-making process, thereby optimizing for latency, risk, and cost to ensure actions are safe, auditable, and executable by construction.
This paper introduces the Independence-Assumption Footprint (IAF) audit protocol to demonstrate that while the NVIDIA Nemotron-Personas-Korea dataset aligns with official marginal demographics, it fails to preserve critical joint-distribution structures across attributes like occupation, age, and gender, thereby establishing that marginal alignment alone is insufficient for ensuring synthetic data fidelity.
This paper proposes a hybrid intelligent navigation framework that integrates an enhanced potential-guided AP-RRT global planner with adaptive weighted Dynamic Window Approach local controllers to achieve robust, efficient, and smooth autonomous navigation in complex dynamic environments.
This paper introduces SAFER-MIND, a novel framework that enhances the safety and interpretability of mental health risk detection in social media by combining multidimensional text integration, calibrated probabilistic modeling, and conformal prediction to distinguish between high-confidence decisions and cases requiring human review, thereby achieving high accuracy while minimizing risky errors.