Gemma-3

IBM Research open-sourced CodeAlchemy, a dataset of nearly 1 trillion tokens of synthetic code training data under Apache 2.0, covering 15 programming languages and including 1.3 million files with ex…

MarkTechPost published a tutorial on training Gemma-3 for structured mathematical reasoning using Tunix GRPO, LoRA adapters, and GSM8K rewards. The workflow includes environment setup, prompt formatti…

Researchers have found that large language models exhibit severe bias and mode collapse when asked to generate random outputs, with models like Qwen3 selecting "Wednesday" 80% of the time when asked f…