Model card
TriSum-J.
TriSum Authorsopen-sourceBART-large distilled from GPT-3.5 with structured rationale
TriSum: Learning Summarization Ability from Large Language Models with Structured Rationale. Zhang et al. NAACL 2024. Joint learning stage variant. Distills GPT-3.5 rationale (aspect-triple-summary structure) into a smaller BART model.
§ 01 · Benchmarks
Every benchmark TriSum-J has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | cnn-/-daily-mail | Computer Vision · Optical Character Recognition | rouge-2 | 22.7% | #5 | 2024-03-15 | source ↗ |
| 02 | cnn-/-daily-mail | Computer Vision · Optical Character Recognition | rouge-1 | 45.7% | #6 | 2024-03-15 | source ↗ |
| 03 | cnn-/-daily-mail | Computer Vision · Optical Character Recognition | rouge-l | 41.9% | #6 | 2024-03-15 | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Papers
1 paper with results for TriSum-J.
- 2024-03-15· Natural Language Processing· 3 results
TriSum: Learning Summarization Ability from Large Language Models with Structured Rationale
§ 05 · Sources & freshness
Where these numbers come from.
arxiv
3
results
3 of 3 rows marked verified.