jz666/llama3-ultrafeedback-templated-ppl-margin-5
收藏Hugging Face2025-12-09 更新2025-12-20 收录
下载链接:
https://hf-mirror.com/datasets/jz666/llama3-ultrafeedback-templated-ppl-margin-5
下载链接
链接失效反馈官方服务:
资源简介:
---
dataset_info:
features:
- name: prompt_id
dtype: string
- name: prompt
dtype: string
- name: all_generated_responses
sequence: string
- name: all_rm_scores
sequence: float64
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected
list:
- name: content
dtype: string
- name: role
dtype: string
- name: ppl_chosen
dtype: float64
- name: nll_chosen
dtype: float64
- name: ppl_rejected
dtype: float64
- name: nll_rejected
dtype: float64
splits:
- name: train
num_bytes: 884573190
num_examples: 59876
- name: test
num_bytes: 28746644
num_examples: 1961
- name: train_correct_q1
num_bytes: 89851929.68100742
num_examples: 6082
- name: train_correct_q2
num_bytes: 89837156.2627764
num_examples: 6081
- name: train_correct_q3
num_bytes: 89837156.2627764
num_examples: 6081
- name: train_correct_q4
num_bytes: 89837156.2627764
num_examples: 6081
- name: train_correct_q5
num_bytes: 89837156.2627764
num_examples: 6081
- name: train_wrong_q1
num_bytes: 86941566.2894983
num_examples: 5885
- name: train_wrong_q2
num_bytes: 86941566.2894983
num_examples: 5885
- name: train_wrong_q3
num_bytes: 86941566.2894983
num_examples: 5885
- name: train_wrong_q4
num_bytes: 86926792.87126729
num_examples: 5884
- name: train_wrong_q5
num_bytes: 86926792.87126729
num_examples: 5884
- name: train_all_q1
num_bytes: 176926456.7345848
num_examples: 11976
- name: train_all_q2
num_bytes: 176911683.3163538
num_examples: 11975
- name: train_all_q3
num_bytes: 176911683.3163538
num_examples: 11975
- name: train_all_q4
num_bytes: 176911683.3163538
num_examples: 11975
- name: train_all_q5
num_bytes: 176911683.3163538
num_examples: 11975
download_size: 1236032568
dataset_size: 2681771863.3431425
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
- split: train_correct_q1
path: data/train_correct_q1-*
- split: train_correct_q2
path: data/train_correct_q2-*
- split: train_correct_q3
path: data/train_correct_q3-*
- split: train_correct_q4
path: data/train_correct_q4-*
- split: train_correct_q5
path: data/train_correct_q5-*
- split: train_wrong_q1
path: data/train_wrong_q1-*
- split: train_wrong_q2
path: data/train_wrong_q2-*
- split: train_wrong_q3
path: data/train_wrong_q3-*
- split: train_wrong_q4
path: data/train_wrong_q4-*
- split: train_wrong_q5
path: data/train_wrong_q5-*
- split: train_all_q1
path: data/train_all_q1-*
- split: train_all_q2
path: data/train_all_q2-*
- split: train_all_q3
path: data/train_all_q3-*
- split: train_all_q4
path: data/train_all_q4-*
- split: train_all_q5
path: data/train_all_q5-*
---
提供机构:
jz666



