---
license: mit
base_model: facebook/bart-base
tags:
- text2text-generation
- style-transfer
- sarcasm
- bart
- seq2seq
language:
- en
pipeline_tag: summarization
---

# BART-Base (Baseline)

Baseline BART-base supervised fine-tuning on sarcastic->non-sarcastic headline pairs. No context enhancement, no RL.

Part of the **Project LLMao** sarcasm style transfer suite (CS4248 Team 14, NUS AY2025/26 S2).
This model rewrites sarcastic news headlines as neutral, factual equivalents while
preserving the underlying meaning.

## Task

**Input**: A sarcastic news headline
**Output**: A non-sarcastic rewrite

Example:
- In: *"Area Man Passionate Defender Of What He Imagines Constitution To Be"*
- Out: *"Man defends his interpretation of the Constitution."*

## Training

- **Base model**: [`facebook/bart-base`](https://huggingface.co/facebook/bart-base) (139M params)
- **Method**: Standard cross-entropy on (sarcastic, non-sarcastic) pairs derived from NHDSD.
- **Dataset**: 10,868 sarcastic->non-sarcastic headline pairs derived from NHDSD
  (News Headlines Dataset for Sarcasm Detection). Non-sarcastic targets were generated
  by an LLM annotator (StepFun Step-3.5 Flash) with cross-validation by Nemotron.
  Split: `sar_to_non (original)`.
- **Input format**: Raw sarcastic headline (no task prefix — BART is not pretrained with prefixes).
- **Generation**: beam search with `num_beams=4`, `max_length=128`.

## Usage

```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

model_id = "SeeYangZhi/BART-Base-Sarcasm-Rewriter"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

headline = "Area Man Passionate Defender Of What He Imagines Constitution To Be"
inputs = tokenizer(headline, return_tensors="pt", truncation=True, max_length=128)
outputs = model.generate(**inputs, max_length=128, num_beams=4)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Evaluation

Evaluated on a 2,857-sample held-out test split alongside 13 other model variants
(BART, T5 baselines, LLaMA 3.2, ablation studies). Metrics include:

| Metric | Direction |
|---|---|
| Hard Flip Rate (% of samples where sarcasm was removed) | higher ↑ |
| Semantic Similarity (all-MiniLM-L6-v2 cosine) | higher ↑ |
| BLEU vs input (lower = more genuine rewriting) | lower ↓ |
| Perplexity (GPT-2) | lower ↓ |
| Normalized edit distance | higher ↑ |
| Paraphrase score (low = real rewriting) | lower ↓ |

Full per-variant numbers are published alongside the Project LLMao webapp.

## Related models

- [`SeeYangZhi/Llama-3.2-1B-Sarcasm-Rewriter`](https://huggingface.co/SeeYangZhi/Llama-3.2-1B-Sarcasm-Rewriter) — instruction-tuned LLaMA variant
- `SeeYangZhi/BART-Base-Sarcasm-Rewriter` — supervised baseline
- `SeeYangZhi/BART-Base-CE-Sarcasm-Rewriter` — context-enhanced SFT
- `SeeYangZhi/BART-Base-RL-Sarcasm-Rewriter` — REINFORCE on top of baseline
- `SeeYangZhi/BART-Base-CE-RL-Sarcasm-Rewriter` — CE + RL (best)

## License

MIT, inheriting from `facebook/bart-base`. The NHDSD dataset is used under its
original research-use terms.