Synthetic conversational personality dataset for the Love Game experiments.
Milestone: 2000000
Combined rows: 3651
Combined tokens: 2166051
Tokenizer used for accounting: HuggingFaceTB/SmolLM2-135M-Instruct
dpo_train.jsonl: 229 rows, 100372 tokens
grpo_prompts.jsonl: 135 rows, 45929 tokens
ppo_prompts.jsonl: 1118 rows, 888979 tokens
reward_model_train.jsonl: 229 rows, 145207 tokens
rl_train.jsonl: 135 rows, 39408 tokens… See the full description on the dataset page:
https://huggingface.co/datasets/sanjuhs/adt-personality-dataset.