Skip to content

Dataset · IlyaGusev

Twice-told conference reviews

IlyaGusev/twicetold

A corpus of scientific conference reviews with LLM-extracted claims and metrics.

What's inside

Contains 527,109 reviews for 164,840 papers from 3 ML conferences and 5 journals. It includes LLM-extracted claims, issue groups, and metrics (MELD, EditLens, GPTTrace, GigaCheck) for English LLM-generated text.

What you can do with it

  • Analyzing review texts and LLM generation quality
  • Training text classifiers and data analysis systems
Tasks
Text classification
Languages
English
Size
100K–1M rows
Downloads this month
773
Likes
2

How to load it

from datasets import load_dataset

ds = load_dataset("IlyaGusev/twicetold")