metadata
language: en
license: apache-2.0
library_name: transformers
pipeline_tag: summarization
tags:
- scihigh-2026
- research-highlight-generation
- scientific-summarization
- text2text-generation
- seq2seq
- bart
base_model: facebook/bart-large-cnn
Origin Task 1 — BART-large-CNN
This is a full-parameter fine-tuned checkpoint of facebook/bart-large-cnn for SciHigh-2026 Task 1. It generates concise research highlights from scientific paper abstracts.
Model Details
- Parameters: 406M trainable
- Framework: Transformers 4.57.6, PyTorch 2.13.0+cu130
- Training Data: 10,000 examples (SciHigh Task 1 MixSub-SciHigh split)
- Hyperparameters: 3 epochs, lr
3e-5, batch size 32, max input/target lengths 1024/320, bfloat16, SDPA attention.
Validation Results (1,985 examples)
| Metric | Score |
|---|---|
| ROUGE-1 | 0.3808 |
| ROUGE-2 | 0.1403 |
| ROUGE-L / Lsum | 0.2503 / 0.2505 |
| METEOR | 0.3114 |
| BERTScore F1 | 0.8721 |
Inference
import torch
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
model_id = "uthayamurthy/origin-task1-bart-large"
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)
model = AutoModelForSeq2SeqLM.from_pretrained(
model_id,
dtype=torch.bfloat16,
attn_implementation="sdpa",
).to("cuda").eval()
abstract = "..."
inputs = tokenizer(abstract, return_tensors="pt", truncation=True, max_length=1024)
inputs = {name: value.to("cuda") for name, value in inputs.items()}
with torch.inference_mode():
generated = model.generate(
**inputs,
max_length=128,
min_length=0,
num_beams=4,
length_penalty=1.0,
no_repeat_ngram_size=3,
early_stopping=True,
)
print(tokenizer.decode(generated[0], skip_special_tokens=True))