GPT-style models
PulseAugur coverage of GPT-style models — every cluster mentioning GPT-style models across labs, papers, and developer communities, ranked by signal.
-
New method estimates language model capacity at 3.6 bits/parameter
Researchers have developed a novel method to quantify the memorization capacity of language models, distinguishing between unintended memorization and generalization. Their findings suggest that GPT-style models possess…
-
New framework unifies in-context learning analysis for causal and masked models
Researchers have developed a unified statistical learning framework to analyze in-context learning (ICL) capabilities in both causal and masked language models. This framework places autoregressive and masked pretrainin…
-
Individual trains GPT-style models on GTX 1050 GPU
An individual details their experience training GPT-style language models on a consumer-grade GTX 1050 GPU with limited VRAM. Initially aiming to train a coding model, they encountered dataset size limitations and shift…