Research map / Large language models / Retrieval and knowledge
CREAK: A Dataset for Commonsense Reasoning over Entity Knowledge
Yasumasa Onoe, Michael JQ Zhang, Eunsol Choi, Greg Durrett
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train on CREAK instead of zero-shot FEVER-KILT | RoBERTa-Large · CREAK test | accuracy: 70.2 -> 80.3 | Improved |
| scale RoBERTa-Large to T5-3b, in-domain | CREAK-trained model · CREAK test | accuracy: 80.3 -> 85.1 | Improved |
| scale RoBERTa-Base to RoBERTa-Large, in-domain | CREAK-trained model · CREAK test | accuracy: 71.6 -> 80.3 | Improved |
| scale RoBERTa-Large to T5-3b, trained on FEVER-KILT | zero-shot FEVER-KILT · CREAK test | accuracy: 70.2 -> 76.7 | Improved |
| pre-train on FEVER-KILT before CREAK finetuning | RoBERTa-Large · CREAK test | accuracy: 80.3 -> 81.1 | No clear effect |
| remove entity mentions from claims (ablation) | RoBERTa-Large trained on CREAK · CREAK test | accuracy: 80.3 -> 67.3 | Hurt |
| add 3 DPR-retrieved Wikipedia passages | RoBERTa-Large trained on CREAK · CREAK test | accuracy: 80.3 -> 84.3 | Improved |
| finetune on FEVER-KILT first (retrieval setting) | RoBERTa-Large+DPR · CREAK test | accuracy: 84.3 -> 86.8 | Improved |
| train on FaVIQ instead of FEVER-KILT zero-shot | RoBERTa-Large · CREAK test | accuracy: 70.2 -> 52.6 | Hurt |
Closest papers by meaning
- Fact-Driven Logical Reasoning for Machine Reading Comprehension (AAAI 2024 · similarity 0.87)
- CORECODE: A Common Sense Annotated Dialogue Dataset with Benchmark Tasks for Chinese Large Language Models (AAAI 2024 · similarity 0.86)
- Exploring the Limits of Sub-Billion Language Model Reasoners with Open Training Recipes (ICLR 2026 · similarity 0.84)
- TACT: Advancing Complex Aggregative Reasoning with Information Extraction Tools (NeurIPS 2024 · similarity 0.84)
- metabench - A Sparse Benchmark of Reasoning and Knowledge in Large Language Models (ICLR 2025 · similarity 0.83)
- PhantomWiki: On-Demand Datasets for Reasoning and Retrieval Evaluation (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
