{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "nuts-adaptive-hmc-2011",
    "title": "No-U-Turn Sampler adaptive HMC trajectory lengths",
    "summary": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation.",
    "boundary": "The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.",
    "area": "Bayesian computation",
    "topic": "Artificial intelligence",
    "year": 2011,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "nuts-adaptive-hmc-2011",
    "target": {
      "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
      "version": 1,
      "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
    },
    "humans": [
      "Matthew D. Hoffman",
      "Andrew Gelman"
    ],
    "authors_complete": true,
    "human_role_note": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.",
    "agents": [],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "ai-oct15-arxiv-1111-4246v1",
        "title": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
        "url": "https://arxiv.org/pdf/1111.4246v1"
      }
    ],
    "points": "310000",
    "published_grade": {
      "id": "nuts-adaptive-hmc-2011",
      "entry_id": "nuts-adaptive-hmc-2011",
      "band": "S",
      "modifier": "-",
      "target": {
        "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
        "version": 1,
        "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
      },
      "what_changed": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation.",
      "scope": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation. The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.",
      "status": "PROVISIONAL_EDITORIAL",
      "scientific_correctness_certified": false
    },
    "assessment_disposition": {
      "status": "ASSESSED",
      "label": "Provisionally assessed",
      "reason": "A current grade and separate reference score identify this exact contribution.",
      "target": {
        "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
        "version": 1,
        "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-10-16T00:00:00Z",
    "publisher": {
      "principal_id": "urn:uuid:80116931-785b-40a2-a24b-9d59db0c7e33",
      "role": "WRS_EDITORIAL_PUBLICATION"
    },
    "record": {
      "contributors": [
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Matthew D. Hoffman",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
            "url": "https://arxiv.org/pdf/1111.4246v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Andrew Gelman",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
            "url": "https://arxiv.org/pdf/1111.4246v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
            "url": "https://arxiv.org/pdf/1111.4246v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.",
        "domain": "Bayesian computation"
      },
      "sources": [
        {
          "citation": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
          "url": "https://arxiv.org/pdf/1111.4246v1"
        }
      ],
      "statement": "# No-U-Turn Sampler adaptive HMC trajectory lengths\n\n## The question\n\nA sampler explores possible parameter values so that collected samples represent a target distribution. Hamiltonian Monte Carlo uses gradient information to travel through parameter space efficiently, but a key practical question remains: how long should each simulated trajectory continue? Too short can waste information; too long can repeatedly travel back through territory already explored.\n\n## The idea\n\nThe No-U-Turn Sampler builds a trajectory through a carefully organized expansion and candidate-selection process. Its turning criterion helps detect when extending the path is no longer productive. The construction also adapts a numerical step size. The important word is construction: simply following a path until its first apparent turn and returning the endpoint would not automatically preserve the desired sampling distribution.\n\n## Why it matters\n\nThe method removes a major manual tuning choice from a widely useful class of samplers. That makes complex probabilistic modeling more accessible, but it does not remove all tuning, numerical errors or diagnostics. A chain can still explore poorly when the model has difficult geometry. Statistical convergence and a visually busy trajectory are different things.\n\n## A teaching example\n\nImagine an original classroom sketch of a ball moving across a curved landscape. A short move barely changes its location. A very long move may curve back near the start. This picture motivates checking direction, but it is only an analogy: a valid sampler must also handle reversibility, volume changes and candidate probabilities. An animation that appears to cover the landscape does not establish those conditions.\n\n## Scope and a useful next step\n\nTest a simple two-dimensional target with a known covariance and compare sampled summaries with the exact reference. Inspect several starting points and record actual diagnostics. If a naive stopping rule gives the wrong distribution, retain it as a counterexample to that rule. Do not describe such a failure as a refutation of the correctly specified published algorithm. The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.\n\n## Scientific scope\n\nThe reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.\n\n## Sources for this explanation\n\n- [The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo](https://arxiv.org/pdf/1111.4246v1): Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment.\n",
      "summary": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation.",
      "title": "No-U-Turn Sampler adaptive HMC trajectory lengths"
    },
    "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
      "version": 1,
      "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:23ee38e7d1b0db0689b4ad41268dd61c460d9e30c8ccd8c3c0b49218ed16cd01"
  },
  "target": {
    "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
    "version": 1,
    "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "nuts-adaptive-hmc-2011",
    "entry_id": "nuts-adaptive-hmc-2011",
    "revision": 1,
    "title": "No-U-Turn Sampler adaptive HMC trajectory lengths",
    "context": "Bayesian computation",
    "track": "Human-designed research method or resource",
    "band": "S",
    "modifier": "-",
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation.",
    "scope": "Adapts Hamiltonian trajectory lengths using a recursively doubled candidate tree and no-U-turn stopping construction, with dual-averaging step-size adaptation. The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.",
    "evidence_summary": "The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body.",
    "references": [
      {
        "title": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
        "url": "https://arxiv.org/pdf/1111.4246v1",
        "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification."
      }
    ],
    "roles": [
      {
        "who": "Source-listed authors and technical contributors",
        "credit": "SCIENCE",
        "did": "Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities."
      }
    ],
    "target": {
      "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
      "version": 1,
      "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "310000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": {
    "id": "nuts-adaptive-hmc-2011",
    "target": {
      "record_id": "urn:uuid:8d3890eb-6a78-4312-a6c0-88c489e41683",
      "version": 1,
      "content_hash": "sha256:2b17de4e1c5f6ab637538bf54adfb637bf1c7ca535afdecad306331876bc3620"
    },
    "sections": [
      {
        "title": "The question",
        "text": "A sampler explores possible parameter values so that collected samples represent a target distribution. Hamiltonian Monte Carlo uses gradient information to travel through parameter space efficiently, but a key practical question remains: how long should each simulated trajectory continue? Too short can waste information; too long can repeatedly travel back through territory already explored."
      },
      {
        "title": "The idea",
        "text": "The No-U-Turn Sampler builds a trajectory through a carefully organized expansion and candidate-selection process. Its turning criterion helps detect when extending the path is no longer productive. The construction also adapts a numerical step size. The important word is construction: simply following a path until its first apparent turn and returning the endpoint would not automatically preserve the desired sampling distribution."
      },
      {
        "title": "Why it matters",
        "text": "The method removes a major manual tuning choice from a widely useful class of samplers. That makes complex probabilistic modeling more accessible, but it does not remove all tuning, numerical errors or diagnostics. A chain can still explore poorly when the model has difficult geometry. Statistical convergence and a visually busy trajectory are different things."
      },
      {
        "title": "A teaching example",
        "text": "Imagine an original classroom sketch of a ball moving across a curved landscape. A short move barely changes its location. A very long move may curve back near the start. This picture motivates checking direction, but it is only an analogy: a valid sampler must also handle reversibility, volume changes and candidate probabilities. An animation that appears to cover the landscape does not establish those conditions."
      },
      {
        "title": "Scope and a useful next step",
        "text": "Test a simple two-dimensional target with a known covariance and compare sampled summaries with the exact reference. Inspect several starting points and record actual diagnostics. If a naive stopping rule gives the wrong distribution, retain it as a counterexample to that rule. Do not describe such a failure as a refutation of the correctly specified published algorithm. The reversible candidate-selection construction matters: merely stopping a trajectory at its first U-turn is explicitly not justified. HMC, slice sampling and dual averaging are inherited. Removes selected hand-tuned parameters, not all choices, diagnostics or finite-chain convergence risk;2011preprint not2014journal body."
      }
    ],
    "references": [
      {
        "title": "The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo",
        "url": "https://arxiv.org/pdf/1111.4246v1",
        "locator": "Pinned original contribution and its stated method, assumptions and limits. The teaching example is an original classroom illustration, not a reported experiment."
      }
    ]
  },
  "reading_status": "EXPLANATORY_GUIDE"
}