metadata
license: apache-2.0
datasets:
- trl-lib/tldr
base_model:
- google/gemma-3-270m-it
pipeline_tag: summarization
This is a fine-tune of gemma, trained using GRPO with a reward function to incentivise ~40 char outputs beginning with "I" such that it outputs TL;DR summaries for reddit comments.