{
  "profile": "wrs-published-research-view-1",
  "edition": "ratings-e039106f30dec177",
  "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS",
  "entry": {
    "id": "ramesh-et-al-unclip-dalle2-2022",
    "title": "Ramesh and colleagues: unCLIP text-to-image generation (DALL·E 2)",
    "summary": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations.",
    "boundary": "This is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.",
    "area": "Generative image models",
    "topic": "Artificial intelligence",
    "year": 2022,
    "lane": "HUMAN",
    "result": "ALGORITHM",
    "package_id": "ramesh-et-al-unclip-dalle2-2022",
    "target": {
      "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
      "version": 1,
      "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
    },
    "humans": [
      "Aditya Ramesh",
      "Prafulla Dhariwal",
      "Alex Nichol",
      "Casey Chu",
      "Mark Chen"
    ],
    "authors_complete": true,
    "human_role_note": "Complete named paper byline. Human research, source-reported system execution and experimental validation are distinct; individual effort shares are not measured.",
    "agents": [],
    "related": [],
    "subresults": [],
    "scientific_assessment": {
      "source_review": "SOURCE_CURATED",
      "scientific_review": "NOT_REVIEWED_BY_WRS",
      "formal_replay": "NOT_RUN",
      "provider_attestation": "NONE"
    },
    "sources": [
      {
        "id": "arxiv-unclip-2204-06125",
        "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
        "url": "https://arxiv.org/abs/2204.06125v1"
      },
      {
        "id": "openai-unclip-2022",
        "title": "Hierarchical text-conditional image generation with CLIP latents — OpenAI",
        "url": "https://openai.com/index/hierarchical-text-conditional-image-generation-with-clip-latents/"
      }
    ],
    "points": "45000",
    "published_grade": {
      "id": "ramesh-et-al-unclip-dalle2-2022",
      "entry_id": "ramesh-et-al-unclip-dalle2-2022",
      "band": "A",
      "modifier": null,
      "target": {
        "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
        "version": 1,
        "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
      },
      "what_changed": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations.",
      "scope": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations. Scope limitation: This is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.",
      "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:2651f8c1-227d-451e-aebc-1ada0c6ed076",
        "version": 1,
        "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
      },
      "edition": "ratings-e039106f30dec177"
    }
  },
  "version": {
    "kind": "PUBLIC_RECORD_VIEW",
    "previous": null,
    "published_at": "2026-09-27T00:00:00Z",
    "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": "Aditya Ramesh",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Prafulla Dhariwal",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Alex Nichol",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Casey Chu",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        },
        {
          "attribution_basis": "SOURCE_CITED",
          "credit": "historical_scientific",
          "entity_kind": "historical_person",
          "name": "Mark Chen",
          "note": "Archive attribution; detailed role assignment remains source-bound",
          "roles": [
            "formulator"
          ],
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        }
      ],
      "evidence": [
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-0",
          "kind": "source_reference",
          "source": {
            "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
            "url": "https://arxiv.org/abs/2204.06125v1"
          }
        },
        {
          "description": "Bibliographic/stated-result source; not a WRS proof verification",
          "id": "source-1",
          "kind": "source_reference",
          "source": {
            "citation": "Hierarchical text-conditional image generation with CLIP latents — OpenAI",
            "url": "https://openai.com/index/hierarchical-text-conditional-image-generation-with-clip-latents/"
          }
        }
      ],
      "origin": "HUMAN_LEGACY",
      "record_profile": "wrs-local-research-notes-0.1",
      "record_type": "research_note",
      "rrp_version": "0.1",
      "scope": {
        "description": "This is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.",
        "domain": "Generative image models"
      },
      "sources": [
        {
          "citation": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
          "url": "https://arxiv.org/abs/2204.06125v1"
        },
        {
          "citation": "Hierarchical text-conditional image generation with CLIP latents — OpenAI",
          "url": "https://openai.com/index/hierarchical-text-conditional-image-generation-with-clip-latents/"
        }
      ],
      "statement": "# Ramesh and colleagues: unCLIP text-to-image generation (DALL·E 2)\n\nIntroduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations.\n\n## Scientific scope\n\nThis is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.\n\n## Reading status\n\nThis is a source-linked WRS catalogue note. A full explanatory guide and independent scientific reconstruction have not been completed for this entry. Read the identified original work for its complete narrative and arguments.\n\n## People and systems\n\nAditya Ramesh; Prafulla Dhariwal; Alex Nichol; Casey Chu; Mark Chen\n\nComplete named paper byline. Human research, source-reported system execution and experimental validation are distinct; individual effort shares are not measured.\n\n\n\n## Original sources\n\n- [Hierarchical Text-Conditional Image Generation with CLIP Latents](https://arxiv.org/abs/2204.06125v1)\n- [Hierarchical text-conditional image generation with CLIP latents — OpenAI](https://openai.com/index/hierarchical-text-conditional-image-generation-with-clip-latents/)\n",
      "summary": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations.",
      "title": "Ramesh and colleagues: unCLIP text-to-image generation (DALL·E 2)"
    },
    "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
    "version": 1,
    "source_ref": {
      "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
      "version": 1,
      "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
    },
    "export_view": "wrs-public-editorial-view-1",
    "public_view_hash": "sha256:b46a1b2d194c2a9b60023f3acf3f0169df37355ee624c733f70ee69fbd8553eb"
  },
  "target": {
    "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
    "version": 1,
    "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
  },
  "published_grade": {
    "profile": "wrs-published-grade-1",
    "edition": "2026-10-28.1",
    "id": "ramesh-et-al-unclip-dalle2-2022",
    "entry_id": "ramesh-et-al-unclip-dalle2-2022",
    "revision": 1,
    "title": "Ramesh and colleagues: unCLIP text-to-image generation (DALL·E 2)",
    "context": "Generative image models",
    "track": "Human-designed AI research method or resource",
    "band": "A",
    "modifier": null,
    "status": "PROVISIONAL_EDITORIAL",
    "assessed_on": "2026-09-27",
    "evidence_cutoff": "2026-09-27",
    "achievement_state": "SUPPORTED_SCOPED_RESULT",
    "what_changed": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations.",
    "scope": "Introduced a two-stage text-conditional image generator that predicts a CLIP image embedding from a caption and decodes an image conditioned on that embedding, enabling image variations and language-guided manipulations. Scope limitation: This is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.",
    "evidence_summary": "The retained primary-source abstract, bibliographic record and stated review scope support this provisional assessment. No independent experimental replication by WRS is claimed. This is the five-author unCLIP paper, not a claim that OpenAI invented diffusion or that every DALL·E 2 product capability was proved in the paper. Quality and diversity comparisons are those reported by the authors.",
    "references": [
      {
        "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents",
        "url": "https://arxiv.org/abs/2204.06125v1",
        "locator": "Abstract and complete bibliographic byline for 2204.06125v1; not an independent proof or experimental replication."
      },
      {
        "title": "Hierarchical text-conditional image generation with CLIP latents — OpenAI",
        "url": "https://openai.com/index/hierarchical-text-conditional-image-generation-with-clip-latents/",
        "locator": "Source-linked contribution and stated scientific limits; not an independent replication."
      }
    ],
    "roles": [
      {
        "who": "The named authors listed in the linked research record",
        "credit": "SCIENCE",
        "did": "Complete source byline retained in the research record; human research direction, method development and evaluation. Individual labor shares are not measured."
      }
    ],
    "target": {
      "record_id": "urn:uuid:2651f8c1-227d-451e-aebc-1ada0c6ed076",
      "version": 1,
      "content_hash": "sha256:7d7be24676b5be828170aaa0d070ee9f006bc6ab41c0d89ecb4f031136aa78e1"
    },
    "scientific_correctness_certified": false,
    "independent_expert_calibration": false,
    "method": "PROPRIETARY",
    "points": "45000"
  },
  "canonical_agent_submission": false,
  "scientific_verification_awarded": false,
  "reader_guide": null,
  "reading_status": "SOURCE_LINKED_NOTE"
}