{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "andrychowicz-et-al-hindsight-replay-2017",
    "title": "Hindsight Experience Replay: relabeling achieved goals",
    "summary": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal.",
    "boundary": "Assumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.",
    "area": "Goal-conditioned reinforcement learning",
    "topic": "Artificial intelligence",
    "year": 2017,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "andrychowicz-et-al-hindsight-replay-2017",
    "target": {
      "record_id": "urn:uuid:28a643bf-d1ce-4c8a-a27b-942b66020477",
      "version": 1,
      "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
    },
    "humans": [
      "Marcin Andrychowicz",
      "Filip Wolski",
      "Alex Ray",
      "Jonas Schneider",
      "Rachel Fong",
      "Peter Welinder",
      "Bob McGrew",
      "Josh Tobin",
      "Pieter Abbeel",
      "Wojciech Zaremba"
    ],
    "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": "cur20261015-arxiv-1707-01495v1",
        "title": "Hindsight Experience Replay",
        "url": "https://arxiv.org/abs/1707.01495v1"
      }
    ],
    "points": "165000",
    "published_grade": {
      "id": "andrychowicz-et-al-hindsight-replay-2017",
      "entry_id": "andrychowicz-et-al-hindsight-replay-2017",
      "band": "A",
      "modifier": "+",
      "target": {
        "record_id": "urn:uuid:28a643bf-d1ce-4c8a-a27b-942b66020477",
        "version": 1,
        "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
      },
      "what_changed": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal.",
      "scope": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal. Assumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.",
      "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:28a643bf-d1ce-4c8a-a27b-942b66020477",
        "version": 1,
        "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
      },
      "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": "Marcin Andrychowicz",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Filip Wolski",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Alex Ray",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Jonas Schneider",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Rachel Fong",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Peter Welinder",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Bob McGrew",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Josh Tobin",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Pieter Abbeel",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Wojciech Zaremba",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Hindsight Experience Replay",
            "url": "https://arxiv.org/abs/1707.01495v1"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "Assumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.",
        "domain": "Goal-conditioned reinforcement learning"
      },
      "sources": [
        {
          "citation": "Hindsight Experience Replay",
          "url": "https://arxiv.org/abs/1707.01495v1"
        }
      ],
      "statement": "# Hindsight Experience Replay: relabeling achieved goals\n\nReplayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal.\n\n## Scientific scope\n\nAssumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.\n\n## People and systems\n\nMarcin Andrychowicz; Filip Wolski; Alex Ray; Jonas Schneider; Rachel Fong; Peter Welinder; Bob McGrew; Josh Tobin; Pieter Abbeel; Wojciech Zaremba\n\n\n\n## Original sources\n\n- [Hindsight Experience Replay](https://arxiv.org/abs/1707.01495v1)\n",
      "summary": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal.",
      "title": "Hindsight Experience Replay: relabeling achieved goals"
    },
    "record_id": "urn:uuid:28a643bf-d1ce-4c8a-a27b-942b66020477",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:28a643bf-d1ce-4c8a-a27b-942b66020477",
      "version": 1,
      "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:870f287e31674020f36473df88abd0307192666a0eed309413ec5c92e5561c43"
  },
  "target": {
    "record_id": "urn:uuid:28a643bf-d1ce-4c8a-a27b-942b66020477",
    "version": 1,
    "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "andrychowicz-et-al-hindsight-replay-2017",
    "entry_id": "andrychowicz-et-al-hindsight-replay-2017",
    "revision": 1,
    "title": "Hindsight Experience Replay: relabeling achieved goals",
    "context": "Goal-conditioned reinforcement learning",
    "track": "Human-designed research method or resource",
    "band": "A",
    "modifier": "+",
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-10-16",
    "evidence_cutoff": "2026-10-16",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal.",
    "scope": "Replayed trajectories with alternative achieved goals and recomputed goal-dependent rewards, allowing off-policy goal-conditioned agents to learn from attempts that missed the requested goal. Assumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.",
    "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. Assumes goals can be represented and achieved goals identified, reward recomputation is available, and goals do not change environment transition dynamics. Not reward-free learning or arbitrary-task hindsight optimization. Unbiased relabeling is not established for arbitrary stochastic settings.",
    "references": [
      {
        "title": "Hindsight Experience Replay",
        "url": "https://arxiv.org/abs/1707.01495v1",
        "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:28a643bf-d1ce-4c8a-a27b-942b66020477",
      "version": 1,
      "content_hash": "sha256:20794f041e6434b091d249a197859663df3da7d8dc33949f12e550f9a5e873c4"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "165000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}