From the model card
Fields below are copied from the tags and counters on the HuggingFace repository facebook/bart-large-cnn at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Pipeline tag
- summarization
- Library
- Transformers
- Framework tags
- PyTorch, TensorFlow, JAX, Rust (candle)
- Weight formats
- safetensors
- License tag
mit— read the license file in the repo before relying on it- Language tags
- English (en)
- Papers cited
- arXiv:1910.13461
- Datasets declared
- cnn_dailymail
- Downloads (HF counter at last fetch)
- 1,279,315
- Likes (HF counter at last fetch)
- 1,611
- Model card
- https://huggingface.co/facebook/bart-large-cnn
Use cases
- Summarizing news articles and long-form text
- Baseline comparison for evaluating newer summarization models
- Extracting key points from reports or documentation
- Legacy summarization pipelines already using BART
Pros
- Strong ROUGE scores on CNN/DailyMail benchmark
- MIT licensed
- Handles up to 1024 tokens of input
- Well-documented and widely used — extensive community examples
Cons
- 1024-token input limit is restrictive for long documents
- Outdated — PEGASUS, PRIMERA, and LLM-based summarization outperform it
- CNN/DailyMail fine-tuning biases toward news article style
- Can produce abstractive errors (hallucinated facts) on dense source text
Tags
transformerspytorchtfjaxrustsafetensorsbarttext2text-generationsummarizationendataset:cnn_dailymailarxiv:1910.13461license:mitmodel-indexendpoints_compatibleregion:usdeploy:azure