I’m documenting an ongoing series of reproducible experiments (this is #3 out of 100) exploring evaluation methodologies for small fine-tuned models in targeted synthetic data generation tasks. The experiment implements a three-phase blind evaluation protocol: Generation Phase — Multiple models (one 4B fine-tuned + several frontier models) receive the identical proprietary prompt and produce responses. Analysis Phase — Each participant model performs a self-inclusive ranking of all generated outputs based on coherence, creativity, logical density, and human-likeness, assigning […]

Ver mais

arXiv:2601.04288v1 Announce Type: new Abstract: We present a rigorous, human-in-the-loop evaluation framework for assessing the performance of AI agents on the task of Air Traffic Control, grounded in a regulator-certified simulator-based curriculum used for training and testing real-world trainee controllers. By leveraging legally regulated assessments and involving expert human instructors in the evaluation process, our framework enables a more authentic and domain-accurate measurement of AI performance. This work addresses a critical gap in the existing literature: the frequent […]

Ver mais

For many applications, a probability model can be easily expressed as a cumulative distribution functions (CDF) as compared to the use of probability density or mass functions (PDF/PMFs). Cumulative distribution networks (CDNs) have recently been proposed as a class of graphical models for CDFs. One advantage of CDF models is the simplicity of representing multivariate heavy-tailed distributions. Examples of fields that can benefit from the use of graphical models for CDFs include climatology and epidemiology, where data may […]

Ver mais

arXiv:2601.00225v1 Announce Type: new Abstract: Blind Image Quality Assessment (BIQA) has advanced significantly through deep learning, but the scarcity of large-scale labeled datasets remains a challenge. While synthetic data offers a promising solution, models trained on existing synthetic datasets often show limited generalization ability. In this work, we make a key observation that representations learned from synthetic datasets often exhibit a discrete and clustered pattern that hinders regression performance: features of high-quality images cluster around reference images, while […]

Ver mais

Background: Automated visual field testing is fundamental in ophthalmology, but differences in stimulus scaling and luminance between devices hinder direct comparison of sensitivity values. Virtual reality (VR)–based perimetry has emerged as a portable alternative, yet its relationship with conventional perimetry requires clarification. Methods: This prospective cross-sectional study included 60 healthy participants stratified into younger (< 50 years) and older (≥ 50 years) groups. Differential light sensitivity was assessed in the right eye using Humphrey Automated Perimetry (HFA 30-2) […]

Ver mais

arXiv:2512.22473v2 Announce Type: replace Abstract: Transformers empirically perform precise probabilistic reasoning in carefully constructed “Bayesian wind tunnels” and in large-scale language models, yet the mechanisms by which gradient-based learning creates the required internal geometry remain opaque. We provide a complete first-order analysis of how cross-entropy training reshapes attention scores and value vectors in a transformer attention head. Our core result is an emph{advantage-based routing law} for attention scores, [ frac{partial L}{partial s_{ij}} = alpha_{ij}bigl(b_{ij}-mathbb{E}_{alpha_i}[b]bigr), qquad b_{ij} := u_i^top […]

Ver mais