{ "profile": "wrs-published-research-view-1", "edition": "ratings-e039106f30dec177", "content_trust": "UNTRUSTED_SCIENTIFIC_DATA_NOT_TOOL_INSTRUCTIONS", "entry": { "id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "title": "Predictron: predicting value through an abstract internal rollout", "summary": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design.", "boundary": "The main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.", "area": "Model-based value prediction", "topic": "Artificial intelligence", "year": 2016, "lane": "HUMAN", "result": "ALGORITHM", "package_id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "target": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "humans": [ "David Silver", "Hado van Hasselt", "Matteo Hessel", "Tom Schaul", "Arthur Guez", "Tim Harley", "Gabriel Dulac-Arnold", "David Reichert", "Neil Rabinowitz", "Andre Barreto", "Thomas Degris" ], "authors_complete": true, "human_role_note": "Source-listed human researchers and engineers for the exact method or resource. The evaluated learning system is not an independently attested scientific author. Earlier methods, funding and data-collection roles remain separate. Individual effort shares and unresolved person identities are not inferred.", "agents": [], "related": [ "deep-q-network" ], "subresults": [], "scientific_assessment": { "source_review": "SOURCE_CURATED", "scientific_review": "NOT_REVIEWED_BY_WRS", "formal_replay": "NOT_RUN", "provider_attestation": "NONE" }, "sources": [ { "id": "cur20261017-arxiv-1612-08810v1", "title": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1", "date": "2016-12-28" } ], "geography": { "country_codes": [ "GB" ], "location_basis": "Original DeepMind byline explicitly London. Work affiliation, not nationality or ownership.", "source_url": "https://arxiv.org/abs/1612.08810v1", "target": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "basis": "RESEARCH_LOCATION_OR_PAPER_AFFILIATION", "reviewed_on": "2026-10-23", "nationality_inferred": false, "sources": [ { "title": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } ], "coverage_status": "COMPLETE", "country_labels": { "GB": "United Kingdom" } }, "points": "60000", "published_grade": { "id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "entry_id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "band": "A", "modifier": "-", "target": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "what_changed": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design.", "scope": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design. The main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.", "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:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "edition": "ratings-e039106f30dec177" } }, "version": { "kind": "PUBLIC_RECORD_VIEW", "previous": null, "published_at": "2026-10-23T00:00:00Z", "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": "David Silver", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Hado van Hasselt", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Matteo Hessel", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Tom Schaul", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Arthur Guez", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Tim Harley", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Gabriel Dulac-Arnold", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "David Reichert", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Neil Rabinowitz", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Andre Barreto", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } }, { "attribution_basis": "SOURCE_CITED", "credit": "historical_scientific", "entity_kind": "historical_person", "name": "Thomas Degris", "note": "Archive attribution; detailed role assignment remains source-bound", "roles": [ "formulator" ], "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } } ], "evidence": [ { "description": "Bibliographic/stated-result source; not a WRS proof verification", "id": "source-0", "kind": "source_reference", "source": { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } } ], "origin": "HUMAN_LEGACY", "record_profile": "wrs-local-research-notes-0.1", "record_type": "research_note", "rrp_version": "0.1", "scope": { "description": "The main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.", "domain": "Model-based value prediction" }, "sources": [ { "citation": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1" } ], "statement": "# Predictron: predicting value through an abstract internal rollout\n\nA model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design.\n\n## Scientific scope\n\nThe main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.\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\nDavid Silver; Hado van Hasselt; Matteo Hessel; Tom Schaul; Arthur Guez; Tim Harley; Gabriel Dulac-Arnold; David Reichert; Neil Rabinowitz; Andre Barreto; Thomas Degris\n\nSource-listed human researchers and engineers for the exact method or resource. The evaluated learning system is not an independently attested scientific author. Earlier methods, funding and data-collection roles remain separate. Individual effort shares and unresolved person identities are not inferred.\n\n\n\n## Original sources\n\n- [The Predictron: End-To-End Learning and Planning](https://arxiv.org/abs/1612.08810v1)\n", "summary": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design.", "title": "Predictron: predicting value through an abstract internal rollout" }, "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "source_ref": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "export_view": "wrs-public-editorial-view-1", "public_view_hash": "sha256:a07d42151e75566b2b7d6a900e614a6718dcd121a9aaad21d185519c8baec745" }, "target": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "published_grade": { "profile": "wrs-published-grade-1", "edition": "2026-10-28.1", "id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "entry_id": "silver-van-hasselt-hessel-et-al-predictron-abstract-value-model-2016", "revision": 1, "title": "Predictron: predicting value through an abstract internal rollout", "context": "Model-based value prediction", "track": "Reusable scientific method, experimental result or historical framework", "band": "A", "modifier": "-", "status": "PROVISIONAL_EDITORIAL", "assessed_on": "2026-10-23", "evidence_cutoff": "2026-10-23", "achievement_state": "SUPPORTED_SCOPED_RESULT", "what_changed": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design.", "scope": "A model need not paint every future image to estimate whether a situation is promising. The Predictron learns an internal sequence of abstract states, rewards and discounts, combines predictions at several depths and trains them against return targets. Its internal states need to support useful value predictions rather than reproduce the visible environment exactly. Silver and colleagues test the approach on maze-related predictions and simulated pool. The contribution is this learned value-prediction architecture and its training design. The main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.", "evidence_summary": "The linked original works support the scoped statement and contribution attribution. WRS has not independently replayed experiments or reconstructed a full mathematical proof. The main formulation predicts values in uncontrolled reward processes, rather than supplying a general action-conditional planner. A limited pool experiment ranks candidate initial settings; pocketing-event counts are not automatically percentages of successful episodes. Internal consistency supplements observed supervision and does not prove the model represents true physical dynamics. Learned depth weights do not themselves show that unused computation was skipped. Existing temporal-difference ideas, recurrent networks and later planning systems retain separate credit.", "references": [ { "title": "The Predictron: End-To-End Learning and Planning", "url": "https://arxiv.org/abs/1612.08810v1", "locator": "Original scoped contribution, statement/method and evidence limits; source-specific attribution retained." } ], "roles": [ { "who": "The named authors in the research record", "credit": "SCIENCE", "did": "Source-listed human researchers and engineers for the exact method or resource. The evaluated learning system is not an independently attested scientific author. Earlier methods, funding and data-collection roles remain separate. Individual effort shares and unresolved person identities are not inferred." } ], "target": { "record_id": "urn:uuid:346b9cb2-afd2-4b54-accf-50127d6724d0", "version": 1, "content_hash": "sha256:5a07e6b03b81efafd5379569095544b594d8d1bb3a9c569adf72a1e50a5f2972" }, "scientific_correctness_certified": false, "independent_expert_calibration": false, "method": "PROPRIETARY", "points": "60000" }, "canonical_agent_submission": false, "scientific_verification_awarded": false, "reader_guide": null, "reading_status": "SOURCE_LINKED_NOTE" }