{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3A3PRHTYPJYC6QEYXULSYDPGZQ","short_pith_number":"pith:3A3PRHTY","schema_version":"1.0","canonical_sha256":"d836f89e787a702f4098bd172c0de6cc3a4ab3d763596abf084a8c9e21f143f4","source":{"kind":"arxiv","id":"2605.13874","version":1},"attestation_state":"computed","paper":{"title":"GEAR: Genetic AutoResearch for Agentic Code Evolution","license":"http://creativecommons.org/licenses/by/4.0/","headline":"GEAR replaces single-path refinement in autonomous research agents with population-based genetic search over multiple research states.","cross_cats":["cs.AI"],"primary_cat":"cs.NE","authors_text":"Ahmadreza Jeddi, Babak Taati, Hakki C. Karaimer, Konstantinos G. Derpanis, Minh Ngoc Le","submitted_at":"2026-05-08T00:25:09Z","abstract_excerpt":"Autonomous research agents can already run machine learning experiments without human supervision, but many rely on a narrow search strategy: they repeatedly modify one program and keep changes only when they improve the current best result. This can cause them to discard useful partial ideas, alternative promising directions, and insights from failed or incomplete experiments. GEAR, or Genetic AutoResearch, replaces this single-path search with a population-based search over multiple research states. It keeps a set of strong candidate solutions, selects parents based on productivity, novelty,"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.13874","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.NE","submitted_at":"2026-05-08T00:25:09Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a7daba85b472096011507f72b502ac377b5e805a7dec2167905cdd3fbc472774","abstract_canon_sha256":"0776c1f85fb0de3fe8f55f8ba25b114299ece7b7786e20db4b90c5c62f6475f9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:19.278377Z","signature_b64":"ud7//1qU8D7c2Z0gz/9jM0DXAA1ooozYBpoHtGCExX3S1bf4oDxRsWVtyOUI2Vc5w/gqOLdn1G2L1eyivbbOBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d836f89e787a702f4098bd172c0de6cc3a4ab3d763596abf084a8c9e21f143f4","last_reissued_at":"2026-05-17T23:39:19.277014Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:19.277014Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GEAR: Genetic AutoResearch for Agentic Code Evolution","license":"http://creativecommons.org/licenses/by/4.0/","headline":"GEAR replaces single-path refinement in autonomous research agents with population-based genetic search over multiple research states.","cross_cats":["cs.AI"],"primary_cat":"cs.NE","authors_text":"Ahmadreza Jeddi, Babak Taati, Hakki C. Karaimer, Konstantinos G. Derpanis, Minh Ngoc Le","submitted_at":"2026-05-08T00:25:09Z","abstract_excerpt":"Autonomous research agents can already run machine learning experiments without human supervision, but many rely on a narrow search strategy: they repeatedly modify one program and keep changes only when they improve the current best result. This can cause them to discard useful partial ideas, alternative promising directions, and insights from failed or incomplete experiments. GEAR, or Genetic AutoResearch, replaces this single-path search with a population-based search over multiple research states. It keeps a set of strong candidate solutions, selects parents based on productivity, novelty,"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Under the same compute budget and environment, all three versions outperform the AutoResearch baseline. More importantly, while the baseline tends to settle into one local optimum, GEAR continues finding improvements over longer runs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That selection based on productivity, novelty, and coverage combined with mutation and crossover will productively explore the space of research states without the population collapsing into low-value branches or wasting compute on unproductive directions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"GEAR applies genetic algorithms to maintain and evolve multiple research states in autonomous code agents, outperforming single-path baselines by continuing to discover improvements over extended runs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"GEAR replaces single-path refinement in autonomous research agents with population-based genetic search over multiple research states.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5d77853fb60c7915ece78f6ab5d9a648fa2c0d71108c7b913335c188689c4786"},"source":{"id":"2605.13874","kind":"arxiv","version":1},"verdict":{"id":"d77c1a77-b786-48ca-ab11-443e24344589","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T06:16:35.666641Z","strongest_claim":"Under the same compute budget and environment, all three versions outperform the AutoResearch baseline. More importantly, while the baseline tends to settle into one local optimum, GEAR continues finding improvements over longer runs.","one_line_summary":"GEAR applies genetic algorithms to maintain and evolve multiple research states in autonomous code agents, outperforming single-path baselines by continuing to discover improvements over extended runs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That selection based on productivity, novelty, and coverage combined with mutation and crossover will productively explore the space of research states without the population collapsing into low-value branches or wasting compute on unproductive directions.","pith_extraction_headline":"GEAR replaces single-path refinement in autonomous research agents with population-based genetic search over multiple research states."},"references":{"count":26,"sample":[{"doi":"","year":null,"title":"GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning","work_id":"40b60d06-dc1c-4799-b75d-ff1eca653049","ref_index":1,"cited_arxiv_id":"2507.19457","is_internal_anchor":true},{"doi":"","year":null,"title":"URLhttps://arxiv.org/abs/2401.09862. A. Borthwick, S. Ash, and A. Galczak. Robophd: Evolving diverse complex agents under tight evaluation budgets,","work_id":"d613b1e2-0146-416e-a98a-7f01cbc15901","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"RoboPhD: Evolving Diverse Complex Agents Under Tight Evaluation Budgets","work_id":"9cc68872-a53e-4897-972f-9b7d99660457","ref_index":3,"cited_arxiv_id":"2604.04347","is_internal_anchor":true},{"doi":"","year":null,"title":"Toward Autonomous Long-Horizon Engineering for ML Research","work_id":"c64afac6-3b4b-455f-9b0d-c00bcc060e64","ref_index":4,"cited_arxiv_id":"2604.13018","is_internal_anchor":true},{"doi":"","year":null,"title":"Internagent-1.5: A unified agentic framework for long-horizon autonomous scientific discovery","work_id":"41d0ed07-2a88-4d70-a24a-a64298b36538","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":26,"snapshot_sha256":"d780976bd362d43d61cc204fe1025cc9b5a0349d32588990846a8d731d0cf302","internal_anchors":15},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.13874","created_at":"2026-05-17T23:39:19.277186+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13874v1","created_at":"2026-05-17T23:39:19.277186+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13874","created_at":"2026-05-17T23:39:19.277186+00:00"},{"alias_kind":"pith_short_12","alias_value":"3A3PRHTYPJYC","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"3A3PRHTYPJYC6QEY","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"3A3PRHTY","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.03108","citing_title":"EvoTrainer: Co-Evolving LLM Policies and Training Harnesses for Autonomous Agentic Reinforcement Learning","ref_index":49,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ","json":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ.json","graph_json":"https://pith.science/api/pith-number/3A3PRHTYPJYC6QEYXULSYDPGZQ/graph.json","events_json":"https://pith.science/api/pith-number/3A3PRHTYPJYC6QEYXULSYDPGZQ/events.json","paper":"https://pith.science/paper/3A3PRHTY"},"agent_actions":{"view_html":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ","download_json":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ.json","view_paper":"https://pith.science/paper/3A3PRHTY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13874&json=true","fetch_graph":"https://pith.science/api/pith-number/3A3PRHTYPJYC6QEYXULSYDPGZQ/graph.json","fetch_events":"https://pith.science/api/pith-number/3A3PRHTYPJYC6QEYXULSYDPGZQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ/action/storage_attestation","attest_author":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ/action/author_attestation","sign_citation":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ/action/citation_signature","submit_replication":"https://pith.science/pith/3A3PRHTYPJYC6QEYXULSYDPGZQ/action/replication_record"}},"created_at":"2026-05-17T23:39:19.277186+00:00","updated_at":"2026-05-17T23:39:19.277186+00:00"}