{
  "schema_version": 1,
  "showcase_date": "2026-10-16",
  "source_run": "2026-10-16 08:00 北京时间 · CCN 快报本地归档",
  "source_record": "DSH-20261013-049 / RSI-Anything",
  "title": "RSI-Anything: Unlocking Recursive Self-Improvement with World Models",
  "arxiv_id": "2610.13912v1",
  "sources": {
    "paper": "https://arxiv.org/abs/2610.13912v1",
    "pdf": "https://arxiv.org/pdf/2610.13912v1",
    "project": "https://phai-labs.com/en/papers/rsi-anything/"
  },
  "input_pdf_md5": "B8CAA1270E4D4ED9B44AF0993DC00BCB",
  "input_pdf_sha256": "cf4675094696a9fb0a2328bcbca333b76ac1b35df4845c874e5d5a6e14f87293",
  "parser": {
    "grobid_version": "0.8.2",
    "grobid_timestamp_utc": "2026-10-16T00:18+0000",
    "ocr": false,
    "reparsed_for_showcase": false
  },
  "xml": {
    "file": "rsi-anything.xml",
    "sha256": "c0a87b5cf229a05cbea132622fcf219f7fb41052e1062723e2a7600262a9356d"
  },
  "validation": {
    "stage": "本轮仅复核已归档最终包的图片封装关系，未重新解析PDF",
    "verified_at": "2026-10-16",
    "valid": true,
    "counts": {
      "graphic_urls": 17,
      "docling_graphic_urls": 17,
      "referenced_images": 17,
      "available_images": 17,
      "missing_references": 0,
      "unindexed_images": 0,
      "duplicate_references": 0,
      "body_refs": 11,
      "body_ref_images": 8,
      "indexed_without_body_ref": 9,
      "body_refs_missing_targets": 0,
      "remaining_grobid_figures": 0,
      "unlinked_body_refs": 8,
      "dangling_visual_ref_targets": 0
    }
  },
  "boundaries": [
    "13 个 picture 指 Docling 导出的图像块，可能含标识或论文图的分块，不等于论文有 13 幅编号 Figure。",
    "8 张有正文关联，9 张仅在索引中；包校验通过不证明所有图表语义与正文位置准确。",
    "本页复用已完成解析包；没有重新解析 PDF，没有公开原始 PDF、中间归档或运行日志。"
  ],
  "images": [
    {
      "file": "images/picture_001.png",
      "xml_id": "docling_picture_001",
      "kind": "figure",
      "page": 1,
      "body_linked": false,
      "caption": "",
      "width": 169,
      "height": 74,
      "sha256": "848401ce96291c8afa0decc7f8a0fe2ea1cee6f14e90dd90becb43cb8606ff83"
    },
    {
      "file": "images/picture_002.png",
      "xml_id": "docling_picture_002",
      "kind": "figure",
      "page": 2,
      "body_linked": true,
      "caption": "Figure 1 | RSI-Anything overview. Building on the cross-domain predictive foundation of JEPA-Anything, RSI-Anything uses world-model predictions to support understanding, planning, and action selection across diverse tasks. Interaction feedback drives policy or workflow updates and conditional world-model adaptation. Component and compatibility checks determine which policy-model pair is retained to guide the next round of interaction and learning, closing the recursive self-improvement loop.",
      "width": 815,
      "height": 323,
      "sha256": "c172616c992bdb93b8952e729b4e6af595977ab099bfc8777a8a1f50033b9fbc"
    },
    {
      "file": "images/picture_003.png",
      "xml_id": "docling_picture_003",
      "kind": "figure",
      "page": 4,
      "body_linked": false,
      "caption": "",
      "width": 801,
      "height": 560,
      "sha256": "32621e011d4a0de5894cc1dead43d2cd6dd1766f77298f9039d33455bb527d74"
    },
    {
      "file": "images/picture_004.png",
      "xml_id": "docling_picture_004",
      "kind": "figure",
      "page": 4,
      "body_linked": true,
      "caption": "Figure 2 | RSI-Anything couples prediction-guided action with retained system updates. a , Building on JEPA-Anything, world-model predictions guide candidate selection for external evaluation. Only the executed candidate (B) yields paired prediction-observation feedback. Feedback supports policy or workflow revision and reliability assessment; world-model adaptation depends on prediction reliability, available data, and budget. Component and compatibility checks determine the retained policy-model pair, which shapes candidates and learning experiences in the next round. b , The shared mechanism is instantiated in code repair, GUI interaction, laboratory planning, and reaction optimization. An end-to-end immune-science application connects workflow evolution, immune world-model construction, analysis with self-correction, and prospective hypothesis generation.",
      "width": 804,
      "height": 303,
      "sha256": "fa94faac38b866d2953bfbd9ae58c62cca0359b5c1485b908040132f788aa8b0"
    },
    {
      "file": "images/picture_005.png",
      "xml_id": "docling_picture_005",
      "kind": "figure",
      "page": 12,
      "body_linked": true,
      "caption": "Figure 3 | Prediction-guided selection and world-model adaptation in code repair. a , Illustrative repair of a non-convergent loop, from candidate patches and predicted outcomes to execution and test feedback. A-C depict example patches; the benchmark uses a fixed coder and four candidates per round. Feedback supports predictor updates in the adaptive configuration. Panels b-d report results on 200 CodeContests tasks across three seeds with up to three feedback rounds. b , Final hidden-test all-pass rate. c , Candidate-ranking accuracy and five-class Brier score for the static verifier and the fixed-world-model configuration. Higher ranking accuracy and lower Brier score indicate better predictive performance. d , Mean online test-suite calls per task. e,f , Final full-test success averaged over three seeds on APPS Introductory (1,000 problems) and LiveCodeBench v6 (1,055 problems), respectively. Fixed WM denotes RSI-Anything with a fixed world model (CURE-WM-Lite); Adaptive WM denotes RSI-Anything with world-model adaptation. Green shades identify world-model-based methods; gray shades identify baselines.",
      "width": 803,
      "height": 1154,
      "sha256": "e10af7b903c9779333878ce18e6e286c8d57e4bb1aa33de5294b64519f771a33"
    },
    {
      "file": "images/picture_006.png",
      "xml_id": "docling_picture_006",
      "kind": "figure",
      "page": 14,
      "body_linked": false,
      "caption": "",
      "width": 794,
      "height": 334,
      "sha256": "0a4fdad659dd23d1b246634aab6a8179fc12700e10f32fe248a96a70337beb5f"
    },
    {
      "file": "images/picture_007.png",
      "xml_id": "docling_picture_007",
      "kind": "figure",
      "page": 14,
      "body_linked": false,
      "caption": "",
      "width": 795,
      "height": 195,
      "sha256": "19409c14817722603ad806c063adcb85d445e9ca50eecd4d6798c605a587bad4"
    },
    {
      "file": "images/picture_008.png",
      "xml_id": "docling_picture_008",
      "kind": "figure",
      "page": 15,
      "body_linked": true,
      "caption": "Figure 5 | Successive policy improvement with RSI-Anything on the 200-task LabHorizon F200 Level-2 benchmark [22]. a , Illustrative planning workflow connecting protocol context, a 4B policy, predictive evaluation of action sequences and parameters, an execution plan, and assessment. Dashed feedback paths depict update routes in the framework; the quantitative comparison below follows retained policy versions. b , Five consecutive versions of the same 4B policy from one self-improvement process. c , Final Score for each version, combining action-sequence similarity and parameter accuracy. The dashed line marks the Actor-Simulator-Selector reference at 45 . 32%. Versions are ordered by development sequence; spacing is schematic.",
      "width": 775,
      "height": 853,
      "sha256": "2196b197c802c2b3ec471108f9b0b86fcd2e98c2d4cbe845bd9644afc49b1dcf"
    },
    {
      "file": "images/picture_009.png",
      "xml_id": "docling_picture_009",
      "kind": "figure",
      "page": 17,
      "body_linked": false,
      "caption": "Reaction-condition optimization with RSI-Anything (green) and Bayesian optimization (gray). ,",
      "width": 745,
      "height": 431,
      "sha256": "24bb48adefeab2e01c869f300e56e57f2513afe3f01b69ec270348b9d8e5f963"
    },
    {
      "file": "images/picture_010.png",
      "xml_id": "docling_picture_010",
      "kind": "figure",
      "page": 17,
      "body_linked": true,
      "caption": "Figure 6 | a Illustrative GC-screen workflow connecting the reaction setup, catalyst/base/solvent options, prediction-guided condition selection, and measured yield. The reaction couples 4-chloroanisole and 4-(trifluoromethyl)phenylboronic acid using a Pd-NHC-DMS precatalyst. b , Best observed GC yield versus cumulative experiment count. c , GC yield after 20 experiments: 93 . 0% for RSI-Anything and 87 . 5% for Bayesian optimization. d , Illustrative flow-screen workflow for substrates 1c and 2b, with Pd(OAc) 2 , ligand, base, and solvent selection followed by LC-MS analysis. e , Mean best observed yield versus cumulative experiment count in the flow LC-MS screen. f , Yield after 40 experiments: 96 . 15% and 82 . 16%, respectively. Brackets indicate gains of 5 . 5 and 13 . 99 percentage points within the corresponding screen. Condition examples and feedback arrows illustrate the framework; curves and bars report the two separate screens.",
      "width": 703,
      "height": 88,
      "sha256": "0ba086a8e40e1c7b574d0ae12fcaa6ac4481b393378e827e80ea0788ebbaf2c6"
    },
    {
      "file": "images/picture_011.png",
      "xml_id": "docling_picture_011",
      "kind": "figure",
      "page": 17,
      "body_linked": false,
      "caption": "",
      "width": 756,
      "height": 157,
      "sha256": "84f36b84c127e5b763283652a3247aee2e3f34d3b138c66bb5e8d2ec59ba4b5e"
    },
    {
      "file": "images/picture_012.png",
      "xml_id": "docling_picture_012",
      "kind": "figure",
      "page": 17,
      "body_linked": false,
      "caption": "",
      "width": 917,
      "height": 114,
      "sha256": "f3f16c705c9d3a5caa8d2c8942bb4fe5653a39f5a1654240b01c9751b2e20cdf"
    },
    {
      "file": "images/picture_013.png",
      "xml_id": "docling_picture_013",
      "kind": "figure",
      "page": 18,
      "body_linked": true,
      "caption": "",
      "width": 816,
      "height": 859,
      "sha256": "68cccfdc0abb2b8a01a758fd8b52c9f1e85fd7bbd791e4bfcd8cb5a8104ba61f"
    },
    {
      "file": "images/table_001.png",
      "xml_id": "docling_table_001",
      "kind": "table",
      "page": 6,
      "body_linked": false,
      "caption": "Table 1 | Fixed settings, update rules, and retained state in a recursive run.",
      "width": 802,
      "height": 434,
      "sha256": "a9d636def45e4600871cbc378fe13fe5b8b8df855e85e8367ce87bf35cd9856c"
    },
    {
      "file": "images/table_002.png",
      "xml_id": "docling_table_002",
      "kind": "table",
      "page": 13,
      "body_linked": false,
      "caption": "Table 2 | Three-round repair on CodeContests [15] under a shared coder and candidate budget. Both RSI-Anything configurations use world-model-guided candidate selection: one keeps the world model fixed (CURE-WM-Lite), and the other additionally adapts it using feedback.",
      "width": 770,
      "height": 208,
      "sha256": "5d62d7ef1477dffdbebc8ad06f6e1e177e546c877a4d7693b62a1555112bcb8d"
    },
    {
      "file": "images/table_003.png",
      "xml_id": "docling_table_003",
      "kind": "table",
      "page": 15,
      "body_linked": true,
      "caption": "Table 3 | LabHorizon benchmark scores for reference systems [22] and RSI-Anything.",
      "width": 689,
      "height": 308,
      "sha256": "af0d3c37bfaa0eb0c6e955e56bc08a6759ea977e91dbfe6251b55a5cd541247c"
    },
    {
      "file": "images/table_004.png",
      "xml_id": "docling_table_004",
      "kind": "table",
      "page": 21,
      "body_linked": true,
      "caption": "Table 4 | Domain bindings and reported update scope. The world model predicts consequences; the policy proposes or selects actions.",
      "width": 811,
      "height": 802,
      "sha256": "420d5d0d2948895e90a53d93048c9f153551e2fb0775c739667f07236c33506d"
    }
  ]
}