#pretraining
#pretraining

[ follow ]

#concept-frequency #zero-shot-performance #large-language-models #openai #machine-learning #anthropic

Artificial intelligence

fromTechRepublic

OpenAI Co-Founder Andrej Karpathy Joins Anthropic in AI Research Shakeup

Andrej Karpathy joined Anthropic to focus on frontier large language model pretraining research and return to R&D.

Artificial intelligence

OpenAI is secretly fast-tracking 'Garlic' to fix ChatGPT's biggest flaws: What we know

OpenAI is developing a new model codenamed Garlic to improve ChatGPT by altering pretraining so smaller models carry knowledge comparable to larger ones.

Artificial intelligence

Are bad incentives to blame for AI hallucinations? | TechCrunch

Large language models produce confident but false statements because pretraining emphasizes next-word prediction and current evaluation incentives reward superficially correct outputs.

#zero-shot-performance

Artificial intelligence

A Close Look at Misalignment in Pretraining Datasets | HackerNoon

Artificial intelligence

What Happens When Your Prompt Is Too Obscure? | HackerNoon

Artificial intelligence

Analyzing the Impact of Pretraining Frequency on Zero-Shot Performance in Multimodal Models | HackerNoon

Artificial intelligence

A Close Look at Misalignment in Pretraining Datasets | HackerNoon

Artificial intelligence

What Happens When Your Prompt Is Too Obscure? | HackerNoon

Artificial intelligence

Analyzing the Impact of Pretraining Frequency on Zero-Shot Performance in Multimodal Models | HackerNoon

more#zero-shot-performance

Artificial intelligence

How Dataset Diversity Impacts AI Model Performance | HackerNoon

Pretraining data diversity significantly influences model performance, particularly in generalization and predictive capabilities.

Artificial intelligence

AI Training Data Has a Long-Tail Problem | HackerNoon

Pretraining datasets exhibit a long-tailed distribution of concept frequencies, impacting performance disparities.

[ Load more ]