Researchers have developed an optimized machine learning framework using ensemble feature selection and ADASYN-based ...
How Qubic's Neuraxon uses evolution to design AI brains: millions of evolved brains, 500 CPU cores, no GPU, tested on ARC-AGI ...
A deep learning framework combining bird swarm optimization with adaptive image segmentation achieved 99.05 percent accuracy ...
Low-rank adaptation fine-tunes a frozen model by learning small low-rank update matrices rather than changing every original parameter. This guide explains the mechanism, trade-offs, evaluation, and ...
GPT-6 Astra edited Qwen’s weights and raised MATH-500 from 7.9% to 48.1% in author-run tests. How weight-space programming works—and its limits.
Salesforce's SRD distills hindsight into foresight, rescuing agent RL when rewards vanish and lifting success from 0% to 60.6 ...
Cross-validation, hyperparameter tuning, threshold selection, AutoML and model explainability (SHAP). Part of urvee-data-lab - my full archive of data science and analytics work, organised by topic.
Instruction tuning uses examples of instructions and desired responses to make a pretrained model more useful, controllable, ...
In reality the same person rates the same movie differently depending on the situation they watched it in — alone vs. with family, weekday vs. weekend, mood at the time. This project implements and ...
Veovesna SLM is an experimental Indonesian-language small language model built from scratch using a custom decoder-only Transformer architecture in PyTorch. The released checkpoint is Run B-Control, ...
A benchmark for temporal reasoning over patient timelines: given a question about when something happened in one patient's record, retrieve the passages that support the correct answer rather than ...