Article Instance

HTTP 200 OK
Allow: GET
Content-Type: application/json
Vary: Accept

{
    "pk": 50097,
    "title": "Efficient Multi-dimensional Optimization in Abstractive Summarization via Mixture-of-Learnable Prompts Tuning",
    "subtitle": null,
    "abstract": "Human beings prefer to read concise summaries that rephrase the exact ideas of a document using novel statements. Consequently, previous works endeavor to coordinate the faithfulness and abstractiveness of automatic summarization, yet this leads to increased computation or data overhead. To address this problem, we propose a novel prompt tuning approach, MoLP, which allocates the optimization of parallel objectives in abstractive summarization to learnable prompts and effectively relieves the cost burden. Inspired by the neural mixture-of-experts model, MoLP learns input-specific expert prompts to optimize saliency awareness, faithfulness, and abstractiveness, respectively, and learns a task-specific router prompt to weigh and polymerize the experts' effects. More importantly, these lightweight prompts are learned from separate tasks, each built upon a heuristic summary of the same document, significantly saving computing costs and improving data utilization. In experiments, we plug MoLP into frozen language models following the classical prompt tuning setting. Extensive evaluations across four benchmark tasks witness the model-generated summaries with simultaneously improved faithfulness and abstractiveness scores. Few-shot learning tests also underscore the advanced generalization of our method.",
    "language": "eng",
    "license": {
        "name": "",
        "short_name": "",
        "text": null,
        "url": ""
    },
    "keywords": [
        {
            "word": "Artificial Intelligence; Linguistics; Machine learning; Natural Language Processing; Neural Networks"
        }
    ],
    "section": "Abstracts with Poster Presentation (accepted as Abstracts)",
    "is_remote": true,
    "remote_url": "https://escholarship.org/uc/item/5443g1kn",
    "frozenauthors": [
        {
            "first_name": "Fengyu",
            "middle_name": "",
            "last_name": "Lu",
            "name_suffix": "",
            "institution": "Peking University",
            "department": ""
        },
        {
            "first_name": "Jiaxin",
            "middle_name": "",
            "last_name": "Duan",
            "name_suffix": "",
            "institution": "Peking University",
            "department": ""
        },
        {
            "first_name": "Junfei",
            "middle_name": "",
            "last_name": "Liu",
            "name_suffix": "",
            "institution": "Peking University",
            "department": ""
        }
    ],
    "date_submitted": null,
    "date_accepted": null,
    "date_published": "2025-01-01T18:00:00Z",
    "render_galley": null,
    "galleys": [
        {
            "label": "PDF",
            "type": "pdf",
            "path": "https://journalpub.escholarship.org/cognitivesciencesociety/article/50097/galley/38059/download/"
        }
    ]
}