LLParallax/collect-Omni-MATH-filtered-gemma-12b
收藏Hugging Face2026-04-01 更新2026-04-12 收录
下载链接：
https://hf-mirror.com/datasets/LLParallax/collect-Omni-MATH-filtered-gemma-12b
下载链接
链接失效反馈官方服务：
资源简介：
---
dataset_info:
features:
- name: episode_id
dtype: string
- name: timestep
dtype: int64
- name: actions
dtype: string
- name: text_actions
dtype: string
- name: raw_text_actions
dtype: string
- name: judge_feedbacks
dtype: string
- name: plans
dtype: string
- name: plan_completed
dtype: 'null'
- name: adaptive_replan
dtype: 'null'
- name: raw_outputs
dtype: string
- name: rewards
dtype: float64
- name: terms
dtype: bool
- name: truncs
dtype: bool
- name: image_paths
dtype: image
- name: state_dumps
dtype: binary
- name: short_term_context
dtype: string
- name: long_term_context
dtype: string
- name: obs_problem
dtype: string
- name: obs_problem_id
dtype: int64
- name: obs_source
dtype: string
- name: obs_solution
dtype: string
- name: obs_answer
dtype: string
- name: obs_attempts
list:
- name: answer
dtype: string
- name: correct
dtype: bool
- name: feedback
dtype: string
- name: turn
dtype: int64
- name: obs_attempts_remaining
dtype: int64
- name: obs_num_attempts
dtype: int64
- name: obs_last_feedback
dtype: string
- name: episode_index
dtype: int64
- name: llm_backend
dtype: string
- name: llm_model_id
dtype: string
- name: llm_tokenizer_id
dtype: string
- name: llm_base_url
dtype: string
splits:
- name: train
num_bytes: 9014027906
num_examples: 238439
download_size: 4316402980
dataset_size: 9014027906
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---
提供机构：
LLParallax
