---
dataset_info:
  features:
  - name: question
    dtype: string
  - name: primed_question
    sequence: string
  - name: greedy_ans
    sequence: string
  - name: p_greed
    sequence: bool
  - name: answer
    list:
    - name: aliases
      sequence: string
    - name: normalized_aliases
      sequence: string
  - name: Category
    dtype: string
  - name: para
    sequence: string
  - name: para_len
    dtype: float64
  - name: is_para
    dtype: bool
  splits:
  - name: full
    num_bytes: 305630848
    num_examples: 21036
  download_size: 29884416
  dataset_size: 305630848
configs:
- config_name: default
  data_files:
  - split: full
    path: data/full-*
task_categories:
- question-answering
---

Dataset for paper “How Much Knowledge Can You Pack into a LoRA Adapter without Harming LLM?” ([https://huggingface.co/papers/2502.14502](https://huggingface.co/papers/2502.14502))
Based on DBpedia dataset

(arxiv.org/abs/2502.14502)