arXiv:2603.28972v1 Announce Type: new Abstract: The large-scale adoption of Large Language Models (LLMs) forces a trade-off between operational cost (OpEx) and data privacy. Current routing frameworks reduce costs but ignore prompt sensitivity, exposing users and institutions to leakage risks towards third-party cloud providers. We formalise the “Inseparability Paradigm”: advanced context management intrinsically coincides with privacy management. We propose a local “Privacy Guard” — a holistic contextual observer powered by an on-premise Small Language Model (SLM) — that performs […]

Ver mais

Where are the engineering productivity and velocity gains we were promised with AI coding tools? AI tools assist in writing half of Google’s code. Microsoft is not far behind at 30%. With so many more lines of code generated by AI, you may wonder: where are the massive engineering productivity improvements? Where are the spades of new features being delivered? Consider this: shovelware apps have not exploded since coding tools emerged. They’ve actually declined. In fact, there’s a growing […]

Ver mais

arXiv:2603.24392v1 Announce Type: new Abstract: Privacy and algorithmic fairness have become two central issues in modern machine learning. Although each has separately emerged as a rapidly growing research area, their joint effect remains comparatively under-explored. In this paper, we systematically study the joint impact of differential privacy and fairness on classification in a federated setting, where data are distributed across multiple servers. Targeting demographic disparity constrained classification under federated differential privacy, we propose a two-step algorithm, namely FDP-Fair. […]

Ver mais

Hi everyone. This is my first question in Reddit, so I do not know if this the place to publish it. I have been trying to train a PPO model to make a Hopper agent “walk”. I have implemented my own version of the PPO algorithm, so that I can modify the architecture more easily. I have done already a huge hyperparameter search (manually done), changed the reward function to an easier and also more complex one, chatted […]

Ver mais

arXiv:2602.16898v1 Announce Type: new Abstract: Task planning for robotic manipulation with large language models (LLMs) is an emerging area. Prior approaches rely on specialized models, fine tuning, or prompt tuning, and often operate in an open loop manner without robust environmental feedback, making them fragile in dynamic settings.We present MALLVi, a Multi Agent Large Language and Vision framework that enables closed loop feedback driven robotic manipulation. Given a natural language instruction and an image of the environment, MALLVi […]

Ver mais

We study personalized multi-agent average reward TD learning, in which a collection of agents interacts with different environments and jointly learns their respective value functions. We focus on the setting where there exists a shared linear representation, and the agents’ optimal weights collectively lie in an unknown linear subspace. Inspired by the recent success of personalized federated learning (PFL), we study the convergence of cooperative single-timescale TD learning in which agents iteratively estimate the common subspace and local […]

Ver mais

Class imbalance induces systematic bias in deep neural networks by imposing a skewed effective class prior. This work introduces the Neural Prior Estimator (NPE), a framework that learns feature-conditioned log-prior estimates from latent representations. NPE employs one or more Prior Estimation Modules trained jointly with the backbone via a one-way logistic loss. Under the Neural Collapse regime, NPE is analytically shown to recover the class log-prior up to an additive constant, providing a theoretically grounded adaptive signal without […]

Ver mais

arXiv:2509.19189v4 Announce Type: replace-cross Abstract: Scaling laws have emerged as a unifying lens for understanding and guiding the training of large language models (LLMs). However, existing studies predominantly focus on the final-step loss, leaving open whether the entire loss dynamics obey similar laws and, crucially, how the learning rate schedule (LRS) shapes them. We address these gaps in a controlled theoretical setting by analyzing stochastic gradient descent (SGD) on a power-law kernel regression model. The key insight is […]

Ver mais

arXiv:2602.20332v1 Announce Type: new Abstract: Advanced reasoning capabilities in Large Language Models (LLMs) have led to more frequent hallucinations; yet most mitigation work focuses on open-source models for post-hoc detection and parameter editing. The dearth of studies focusing on hallucinations in closed-source models is especially concerning, as they constitute the vast majority of models in institutional deployments. We introduce QueryBandits, a model-agnostic contextual bandit framework that adaptively learns online to select the optimal query-rewrite strategy by leveraging an […]

Ver mais

Learn why AI agents become expensive in production and how ReAct loops, context growth, and poor routing wreck unit economics.

Ver mais