{
  "schema_version": 1,
  "title": "Reproduction: Adaptive Estimation and Inference in Semi-parametric Heterogeneous Clustered Multitask Learning via Neyman Orthogonality",
  "emoji": "🧮",
  "space_id": "snaykey/repro-profiling-irrational-agent",
  "paper": {
    "openreview_id": "5hDvooOKUP"
  },
  "tags": [
    "icml2026-repro",
    "paper-5hDvooOKUP"
  ],
  "updated_at": "2026-07-31T00:00:00Z",
  "root": {
    "slug": "index",
    "title": "Reproduction: Adaptive Estimation and Inference in Semi-parametric Heterogeneous Clustered Multitask Learning via Neyman Orthogonality",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "executive-summary",
        "title": "Executive summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "claim-1-thm35-cluster-recovery",
        "title": "The adaptive orthogonal multitask estimator achieves exact recovery of the latent task clustering with high probability, established via Theorem 3.5 (Section 3, Cluster Recovery).",
        "file": "pages/claim-1-thm35-cluster-recovery/page.md",
        "children": []
      },
      {
        "slug": "claim-2-thm35-pooled-rate",
        "title": "For tasks in cluster k with pooled sample size N_k, the estimator attains the rate ||θ̂_j − θ*_j||_2 = O_P(N_k^{-1/2}), matching pooled parametric convergence (Theorem 3.5, Section 3).",
        "file": "pages/claim-2-thm35-pooled-rate/page.md",
        "children": []
      },
      {
        "slug": "claim-3-thm36-asymptotic-normality",
        "title": "√N_k(θ̂_j − θ*_j) is asymptotically normal with covariance matching the oracle estimator that knows the true clustering in advance (Theorem 3.6, Section 3).",
        "file": "pages/claim-3-thm36-asymptotic-normality/page.md",
        "children": []
      },
      {
        "slug": "claim-4-thm37-38-within-cluster-heterogeneity",
        "title": "Extensions permit within-cluster heterogeneity bounded by ξ_k = O(N_k^{-1/2}) while preserving the estimation guarantees (Theorems 3.7–3.8, Section 3).",
        "file": "pages/claim-4-thm37-38-within-cluster-heterogeneity/page.md",
        "children": []
      },
      {
        "slug": "claim-5-sec44-simulations-ari",
        "title": "In simulations across three models (PLM, ATE, DID) and separation levels δ ∈ {1/3, 2/3, 1}, the method achieves Adjusted Rand Index (ARI) near 1, outperforming competing clustering approaches (Section 4.4).",
        "file": "pages/claim-5-sec44-simulations-ari/page.md",
        "children": []
      },
      {
        "slug": "claim-6-sec5-recs-real-data",
        "title": "On real electricity price elasticity data from 51 US states, the method recovers three clusters, e.g. Virginia at −1.138 ± 0.189 versus a 46-state cluster at −0.221 ± 0.009 (Table 1, Section 5).",
        "file": "pages/claim-6-sec5-recs-real-data/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "Conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  }
}