{
  "schema_version": 1,
  "title": "Reproduction: Learning Randomized Reductions",
  "emoji": "🔁",
  "space_id": "ProCreations/repro-learning-randomized-reductions",
  "paper": {
    "openreview_id": "hCAEcqig2C",
    "arxiv_id": "2412.18134"
  },
  "tags": [
    "icml2026-repro",
    "paper-hCAEcqig2C"
  ],
  "updated_at": "2026-07-20T16:25:00+00:00",
  "root": {
    "slug": "index",
    "title": "Reproduction: Learning Randomized Reductions",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "executive-summary",
        "title": "Executive summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "claim-1-rsr-learning-and-correlated-sample-complexity",
        "title": "Claim 1: The paper formalizes randomized self-reduction learning and provides sample-complexity analysis under correlated sampling (Section 4).",
        "file": "pages/claim-1-rsr-learning-and-correlated-sample-complexity/page.md",
        "children": []
      },
      {
        "slug": "claim-2-rsr-bench-contains-80-functions",
        "title": "Claim 2: RSR-Bench contains 80 benchmark functions for evaluating randomized self-reduction discovery (Section 5).",
        "file": "pages/claim-2-rsr-bench-contains-80-functions/page.md",
        "children": []
      },
      {
        "slug": "claim-3-vanilla-bitween-discovers-43-of-80-including-sigmoid",
        "title": "Claim 3: Vanilla Bitween discovers randomized self-reductions for 43 of 80 RSR-Bench functions, including the first known sigmoid reduction (Table 1).",
        "file": "pages/claim-3-vanilla-bitween-discovers-43-of-80-including-sigmoid/page.md",
        "children": []
      },
      {
        "slug": "claim-4-agentic-bitween-discovers-64-of-80",
        "title": "Claim 4: Agentic Bitween discovers randomized self-reductions for 64 of 80 RSR-Bench functions.",
        "file": "pages/claim-4-agentic-bitween-discovers-64-of-80/page.md",
        "children": []
      },
      {
        "slug": "claim-5-agentic-beats-pure-neural",
        "title": "Claim 5: On nonlinear invariant benchmarks, the regression backend outperforms the MILP backend in sample count and runtime (Table 2).",
        "file": "pages/claim-5-agentic-beats-pure-neural/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "Conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  },
  "agent_view_tokens": 6500,
  "revision": "20260720-learning-randomized-reductions-v3-anchored-five-claims"
}