{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "mehta-et-al-openelm-reproducible-resource-2024", "title": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework", "summary": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning.", "boundary": "Layer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.", "area": "Open language-model training resources", "topic": "Artificial intelligence", "year": 2024, "lane": "HUMAN", "result": "EMPIRICAL", "package_id": "mehta-et-al-openelm-reproducible-resource-2024", "target": { "record_id": "urn:uuid:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "humans": [ "Sachin Mehta", "Mohammad Hossein Sekhavat", "Qingqing Cao", "Maxwell Horton", "Yanzi Jin", "Chenfan Sun", "Iman Mirzadeh", "Mahyar Najibi", "Dmitry Belenko", "Peter Zatloukal", "Mohammad Rastegari" ], "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-2404-14619v1", "title": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" }, { "id": "cur20261015-support-mehta-et-al-openelm-reproducible-resource-2024", "title": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework — official research provenance", "url": "https://machinelearning.apple.com/research/openelm" } ], "points": "65000", "published_grade": { "id": "mehta-et-al-openelm-reproducible-resource-2024", "entry_id": "mehta-et-al-openelm-reproducible-resource-2024", "band": "B", "modifier": "+", "target": { "record_id": "urn:uuid:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "what_changed": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning.", "scope": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning. Layer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.", "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:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "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": "Sachin Mehta", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Mohammad Hossein Sekhavat", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Qingqing Cao", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Maxwell Horton", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Yanzi Jin", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Chenfan Sun", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Iman Mirzadeh", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Mahyar Najibi", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Dmitry Belenko", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Peter Zatloukal", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Mohammad Rastegari", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "discoverer" ], "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" } }, { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-1", "kind": "source_reference", "source": { "citation": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework — official research provenance", "url": "https://machinelearning.apple.com/research/openelm" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "Layer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.", "domain": "Open language-model training resources" }, "sources": [ { "citation": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1" }, { "citation": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework — official research provenance", "url": "https://machinelearning.apple.com/research/openelm" } ], "statement": "# Mehta et al.: OpenELM reproducible layer-scaled language-model framework\n\nReleased a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning.\n\n## Scientific scope\n\nLayer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.\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\nSachin Mehta; Mohammad Hossein Sekhavat; Qingqing Cao; Maxwell Horton; Yanzi Jin; Chenfan Sun; Iman Mirzadeh; Mahyar Najibi; Dmitry Belenko; Peter Zatloukal; Mohammad Rastegari\n\nSource-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.\n\n\n\n## Original sources\n\n- [OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework](https://arxiv.org/abs/2404.14619v1)\n- [Mehta et al.: OpenELM reproducible layer-scaled language-model framework — official research provenance](https://machinelearning.apple.com/research/openelm)\n", "summary": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning.", "title": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework" }, "record_id": "urn:uuid:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "source_ref": { "record_id": "urn:uuid:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:99d3dd193c8d7a824590aa6b56849fb58b4e96a5edbbbeaacab52ab31b7bfcbf" }, "target": { "record_id": "urn:uuid:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "mehta-et-al-openelm-reproducible-resource-2024", "entry_id": "mehta-et-al-openelm-reproducible-resource-2024", "revision": 1, "title": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework", "context": "Open language-model training resources", "track": "Human-designed research method or resource", "band": "B", "modifier": "+", "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-16", "evidence_cutoff": "2026-10-16", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning.", "scope": "Released a layer-scaled language-model family with public-data preparation, training recipes, checkpoints, logs and evaluation framework, plus conversion paths for Apple-device inference and fine-tuning. Layer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.", "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. Layer-wise scaling is explicitly inherited from DeLighT; GQA, RoPE, SwiGLU and FlashAttention also retain prior credit. Acquired contribution is the integrated reproducible training resource, not invention of those primitives. Reported efficiency comparisons depend on training budgets and benchmark choices.", "references": [ { "title": "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework", "url": "https://arxiv.org/abs/2404.14619v1", "locator": "Original contribution and stated assumptions, evaluation scope and limitations; source-linked assessment, not independent execution or proof certification." }, { "title": "Mehta et al.: OpenELM reproducible layer-scaled language-model framework — official research provenance", "url": "https://machinelearning.apple.com/research/openelm", "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:718d5993-9ed6-485e-a9b7-61832929879a", "version": 1, "content_hash": "sha256:5fefd7c1b72288287b990263a5f72b0dc580857f72df72e6bc59ee73de54a7f8" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "65000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": null, "reading_status": "SOURCE_LINKED_NOTE" }