[
    {
        "id": "osp-17594",
        "type": "article-journal",
        "title": "SearchJev: A Fast and Calibrated System-1 Model for Search Agents",
        "author": [
            {
                "family": "Cao",
                "given": "Congfeng"
            },
            {
                "family": "Zuo",
                "given": "Lipeng"
            },
            {
                "family": "Papakostas",
                "given": "Konstantinos"
            },
            {
                "family": "Xu",
                "given": "Qiwei"
            },
            {
                "family": "Xu",
                "given": "Songwei"
            },
            {
                "family": "Zhou",
                "given": "Lun"
            },
            {
                "family": "Ren",
                "given": "Zhaochun"
            },
            {
                "family": "Lyu",
                "given": "Yougang"
            },
            {
                "family": "Yan",
                "given": "Xiaohui"
            }
        ],
        "URL": "https://omanscience.com/en/articles/searchjev-a-fast-and-calibrated-system-1-model-for-search-agents",
        "language": "en",
        "issued": {
            "date-parts": [
                [
                    2026
                ]
            ]
        },
        "abstract": "Search agents repeatedly make short decisions about relevance, evidence sufficiency, and search actions. Using generative language models for these decisions introduces latency and unreliable confidence. We present SearchJev, a fast and calibrated System-1 model that separates search decisions from System-2 reasoning and generation. Given a search state and a decision schema, SearchJev directly scores legal options without autoregressive output generation. We propose Soft-Label Learning for Calibrated Decisions (SLCD) to learn decision probabilities from uncertain supervision and calibrate their confidence. In a dual-system search agent, SearchJev handles short decisions and delegates uncertain judgments to System 2, which retains planning, query generation, and answer composition. We also introduce SearchDecision-Bench, a benchmark unifying six types of search decisions for training and evaluation. On SearchDecision-Bench, SEARCHJEV improves decision quality over same-size Qwen3.5 autoregressive models, achieves 5.2-5.3 times faster decisions, and reduces average expected calibration error by 41-74%. On BrowseComp-Plus, the dual-system agents achieve a 3.7-4.7 times speedup in active search time while improving answer accuracy from 45% to up to 54%."
    }
]