[
    {
        "id": "osp-22941",
        "type": "article-journal",
        "title": "PULSE: Identifying Demonstration-Utility Features with Sparse Autoencoders",
        "author": [
            {
                "family": "Hao",
                "given": "Chenduo"
            },
            {
                "family": "Gao",
                "given": "Chuanbao"
            },
            {
                "family": "Zeng",
                "given": "Pinjun"
            },
            {
                "family": "Zhu",
                "given": "Jingze"
            },
            {
                "family": "Liu",
                "given": "Chonghan"
            },
            {
                "family": "Liu",
                "given": "Zidong"
            },
            {
                "family": "Yang",
                "given": "Xu"
            }
        ],
        "URL": "https://omanscience.com/en/articles/pulse-identifying-demonstration-utility-features-with-sparse-autoencoders",
        "language": "en",
        "issued": {
            "date-parts": [
                [
                    2026
                ]
            ]
        },
        "abstract": "In-context learning is highly sensitive to demonstration choice, yet most methods select demonstrations using external query-demonstration similarity. Such criteria can miss model-specific signals: Similar demonstrations may activate different internal features and downstream behaviors. We introduce PULSE (Paired Utility Localization over Sparse Encodings), an SAE-based framework for identifying model-internal features associated with demonstration utility and using them for demonstration selection. Using a small labeled discovery set, PULSE samples candidate demonstration sets, measures their zero-shot-relative utility under the target model, and scores SAE features by how their activation differences align with utility differences. The top positive and negative coordinates form a sparse utility-localization vector. We use this vector in two complementary ways: as a signed score for controlled complete-set ranking, and as PULSE-Retriever, which converts its magnitude into a feature-relevance mask for scalable pool-scale retrieval. Across classification, generation, and reasoning benchmarks, PULSE-Retriever improves over the strongest baseline by 2-3 accuracy points, 0.6-0.9 BLEU-4, and 3.2 exact-match points, respectively, while controlled ranking validates the identified features encode a predictive set-level utility signal. Feature inspection and cross-dataset experiments suggest that the identified features capture task-relevant, dataset-conditioned patterns, yet retain utility signals that partially transfer across datasets. Our code is available at https://github.com/aohenuo/PULSE."
    }
]