---
dataset_info:
  features:
  - name: doi/arxiv_id
    dtype: string
  - name: title
    dtype: string
  - name: paper_category
    dtype: string
  - name: error_category
    dtype: string
  - name: error_location
    dtype: string
  - name: error_severity
    dtype: string
  - name: error_annotation
    dtype: string
  - name: paper_content
    list:
    - name: image_url
      struct:
      - name: url
        dtype: string
    - name: text
      dtype: string
    - name: type
      dtype: string
  - name: error_local_content
    list:
    - name: image_url
      struct:
      - name: url
        dtype: string
    - name: text
      dtype: string
    - name: type
      dtype: string
  - name: __index_level_0__
    dtype: int64
  splits:
  - name: train
    num_bytes: 58231756
    num_examples: 68
  download_size: 55816319
  dataset_size: 58231756
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
language:
- en
size_categories:
- n<1K
---


---

# SPOT

> Preprocessed Contents of  **Scientific Paper ErrOr DeTection** (SPOT)  
> *SPOT contains 83 papers and 91 human-validated errors to test academic verification capabilities.*  
> *This repo contains preprocessed contents of 62 manuscripts with share-permissive licenses.*  

## 📖 Overview

This repository holds the **full paper files** (parsed Markdown, and base64 encodings of extracted figures) for the subset of SPOT manuscripts that are openly licensed. Combined with the annotations in [SPOT-MetaData](https://github.com/<org>/SPOT-MetaData), you can run end-to-end evaluations of LLMs on multi-modal academic error detection.

> **Benchmark at a glance**
>
> * **83** published manuscripts
> * **91** confirmed errors (errata or retractions)
> * **10** scientific domains (Math, Physics, Biology, …)
> * **6** error types (Equation/Proof, Fig-duplication, Data inconsistency, …)
> * Average paper length: \~12 000 tokens & 18 figures

> **Included**  
> - **62** open-access papers (CC-BY or equivalent)  
> - High-fidelity **Markdown** conversions of each PDF  
> - **base64 encoding** of every figure, table, and equation  
> - All in **openai api** format. 

> **Excluded**  
> - Paywalled or proprietary manuscripts (cannot be redistributed)


## 📋 Column Descriptions

Each row in `annotations/errors.csv` contains the following fields:

* **`doi/arxiv_id`**:
  The paper’s DOI (journal) or arXiv identifier.
* **`title`**:
  Full title of the manuscript.
* **`paper_category`**:
  Scientific domain of the paper, one of:
  Mathematics, Physics, Biology, Chemistry, Materials Science, Medicine, Environmental Science, Engineering, Computer Science, Multidisciplinary.
* **`error_category`**:
  Type of error, one of:

  * `Equation/Proof`
  * `Figure duplication`
  * `Data inconsistency`
  * `Experiment setup`
  * `Reagent identity`
  * `Statistical reporting`
* **`error_location`**:
  Where the error appears (e.g., Figure 2, Equation (5), Section 3.1, Table 4).
* **`error_severity`**:
  Indicates whether the issue led to an `Erratum` correction or a `Retraction`.
* **`error_annotation`**:
  Written summary describing the error.
* **`paper_content`**:
  Processed content of the full paper (text in markdown, images in base64 encodings).
* **`error_local_content`**:
  Extracted snippet around the error—paragraph, caption, or equation block used in experiments in Appendix B.1.
  
## 📜 License & Copyright
> SPOT code: CC-BY-4.0  
> Individual papers & Processed Contents: distributed under their original licenses.