arXiv:2601.08874v1 Announce Type: new Abstract: GenAI systems are increasingly used for drafting, summarisation, and decision support, offering substantial gains in productivity and reduced cognitive load. However, the same natural language fluency that makes these systems useful can also blur the boundary between tool and companion. This boundary confusion may encourage some users to experience GenAI as empathic, benevolent, and relationally persistent. Emerging reports suggest that some users may form emotionally significant attachments to conversational agents, in some cases […]

Ver mais

Rapid advances in artificial intelligence are increasing the rate and steepness of informational and economic gradients experienced by human systems, challenging traditional models of adaptation based on stable identities, static optimization, and long-term professional blueprints. This study proposes a unified dynamical framework connecting thermodynamic entropy, information-theoretic entropy, and a formally defined entropy of the self through a shared stochastic gradient-flow model. Drawing on Langevin dynamics and simulated annealing, physical relaxation, probabilistic learning, and human identity formation are treated […]

Ver mais

Security analysis of learning algorithms is gaining increasing importance, especially since they have become target of deliberate obstruction in certain applications. Some security-hardened algorithms have been previously proposed for supervised learning; however, very little is known about the behavior of anomaly detection methods in such scenarios. In this contribution, we analyze the performance of a particular method—online centroid anomaly detection—in the presence of adversarial noise. Our analysis addresses three key security-related issues: derivation of an optimal attack, analysis […]

Ver mais

Reinforcement learning (RL) agents often suffer from high sample complexity in sparse or delayed reward settings due to limited prior structure. Large language models (LLMs) can provide subgoal decompositions, plausible trajectories, and abstract priors that facilitate early learning. However, heavy reliance on LLM supervision introduces scalability constraints and dependence on potentially unreliable signals. We propose MIRA (Memory-Integrated Reinforcement Learning Agent), which incorporates a structured, evolving memory graph to guide early training. The graph stores decision-relevant information, including trajectory […]

Ver mais

Hey everyone, Meta-reviews just dropped. My paper got two weak rejects and a borderline accept (got dinged for missing some VLM baselines), but the AC recommended it to the new “Findings” track after the AC triplet meeting (not sure what this is). For context, I’m a solo undergrad working entirely without a supervisor. I don’t have a PI or a lab to ask about how this stuff works, so my only source of info is whatever I can […]

Ver mais

arXiv:2601.21058v1 Announce Type: new Abstract: Ising machines have emerged as accelerators for combinatorial optimization. To enable practical deployment, this work aims to reduce time-to-solution by addressing three challenges: (1) hardware topology, (2) spin selection and update algorithms, and (3) scalable coupling-coefficient precision. Restricted topologies require minor embedding; naive parallel updates can oscillate or stall; and limited precision can preclude feasible mappings or degrade solution quality. This work presents Snowball, a digital, scalable, all-to-all coupled Ising machine that integrates […]

Ver mais

I came across this article on data design patterns and found it grounded in real system behavior rather than tools. It walks through patterns that show up when supporting ML and AI workloads at scale. After reading this , I was curious to hear from others here: which patterns you rely on most, which ones failed under scale and patterns you think are overused. I am keen on hearing more about failures and lessons learned than success stories […]

Ver mais