{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DEP2WS5QZP5J5BCIYIBNQY22QI","short_pith_number":"pith:DEP2WS5Q","schema_version":"1.0","canonical_sha256":"191fab4bb0cbfa9e8448c202d8635a823bfee45d5626aae23849ae4cf18880a7","source":{"kind":"arxiv","id":"2604.07303","version":2},"attestation_state":"computed","paper":{"title":"Robots that learn to evaluate models of collective behavior","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A robotic fish distinguishes neural network models of fish behavior as more accurate than rule-based models through closed-loop interactions.","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Andreas Gerken, David Bierbach, Jens Krause, Mathis Hocke, Tim Landgraf","submitted_at":"2026-04-08T17:11:29Z","abstract_excerpt":"Understanding and modeling animal behavior is essential for studying collective motion, decision-making, and bio-inspired robotics. Yet, evaluating the accuracy of behavioral models still often relies on offline comparisons to static trajectory statistics. Here we introduce a reinforcement-learning-based framework that uses a biomimetic robotic fish (RoboFish) to evaluate computational models of live fish behavior through closed-loop interaction. We trained policies in simulation using four distinct fish models-a simple constant-follow baseline, two rule-based models, and a biologically ground"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2604.07303","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2026-04-08T17:11:29Z","cross_cats_sorted":[],"title_canon_sha256":"5ed74519ed329a9a0390f8b49180d0f9d5fbe0382632f3847530e8f0024d5ea9","abstract_canon_sha256":"db1b51c24e8513b6eea18e5ab024469c52c171a9aaa4c9f6ef6d8ab8640047be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:43.091182Z","signature_b64":"LdtBoRqlbLhyoc2UYzQI0z0Fb+VZdr+qsH1+7bo4uskNxLap6rtkxHsYwFPHKnS/eIl/HRa3SNslImgaK/fZBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"191fab4bb0cbfa9e8448c202d8635a823bfee45d5626aae23849ae4cf18880a7","last_reissued_at":"2026-05-20T02:05:43.090292Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:43.090292Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robots that learn to evaluate models of collective behavior","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A robotic fish distinguishes neural network models of fish behavior as more accurate than rule-based models through closed-loop interactions.","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Andreas Gerken, David Bierbach, Jens Krause, Mathis Hocke, Tim Landgraf","submitted_at":"2026-04-08T17:11:29Z","abstract_excerpt":"Understanding and modeling animal behavior is essential for studying collective motion, decision-making, and bio-inspired robotics. Yet, evaluating the accuracy of behavioral models still often relies on offline comparisons to static trajectory statistics. Here we introduce a reinforcement-learning-based framework that uses a biomimetic robotic fish (RoboFish) to evaluate computational models of live fish behavior through closed-loop interaction. We trained policies in simulation using four distinct fish models-a simple constant-follow baseline, two rule-based models, and a biologically ground"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The neural network-based fish model exhibited the smallest gap across goal-reaching performance and most other metrics, indicating higher behavioral fidelity than conventional rule-based models under this benchmark. More importantly, this separation shows that the proposed evaluation can quantitatively distinguish candidate models under matched closed-loop conditions.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That policies trained in simulation to guide a virtual fish to goals transfer sufficiently well to the physical RoboFish hardware and that the chosen behavioral metrics (goal-reaching, inter-individual distances, wall interactions, alignment) are sufficient to reveal meaningful differences in model fidelity.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A robotic fish uses RL policies to interact with live fish and ranks behavioral models by Wasserstein distance between simulated and real distributions of metrics such as goal-reaching performance and alignment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A robotic fish distinguishes neural network models of fish behavior as more accurate than rule-based models through closed-loop interactions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"51e42cee4ab6b60fcb89a60aa6624f8ecd76934bb893977311c157577e352d25"},"source":{"id":"2604.07303","kind":"arxiv","version":2},"verdict":{"id":"dbc9bb1e-a55d-4aad-963a-48d82131ed1a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T17:47:07.594410Z","strongest_claim":"The neural network-based fish model exhibited the smallest gap across goal-reaching performance and most other metrics, indicating higher behavioral fidelity than conventional rule-based models under this benchmark. More importantly, this separation shows that the proposed evaluation can quantitatively distinguish candidate models under matched closed-loop conditions.","one_line_summary":"A robotic fish uses RL policies to interact with live fish and ranks behavioral models by Wasserstein distance between simulated and real distributions of metrics such as goal-reaching performance and alignment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That policies trained in simulation to guide a virtual fish to goals transfer sufficiently well to the physical RoboFish hardware and that the chosen behavioral metrics (goal-reaching, inter-individual distances, wall interactions, alignment) are sufficient to reveal meaningful differences in model fidelity.","pith_extraction_headline":"A robotic fish distinguishes neural network models of fish behavior as more accurate than rule-based models through closed-loop interactions."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.07303/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"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":"2604.07303","created_at":"2026-05-20T02:05:43.090455+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.07303v2","created_at":"2026-05-20T02:05:43.090455+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.07303","created_at":"2026-05-20T02:05:43.090455+00:00"},{"alias_kind":"pith_short_12","alias_value":"DEP2WS5QZP5J","created_at":"2026-05-20T02:05:43.090455+00:00"},{"alias_kind":"pith_short_16","alias_value":"DEP2WS5QZP5J5BCI","created_at":"2026-05-20T02:05:43.090455+00:00"},{"alias_kind":"pith_short_8","alias_value":"DEP2WS5Q","created_at":"2026-05-20T02:05:43.090455+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI","json":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI.json","graph_json":"https://pith.science/api/pith-number/DEP2WS5QZP5J5BCIYIBNQY22QI/graph.json","events_json":"https://pith.science/api/pith-number/DEP2WS5QZP5J5BCIYIBNQY22QI/events.json","paper":"https://pith.science/paper/DEP2WS5Q"},"agent_actions":{"view_html":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI","download_json":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI.json","view_paper":"https://pith.science/paper/DEP2WS5Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.07303&json=true","fetch_graph":"https://pith.science/api/pith-number/DEP2WS5QZP5J5BCIYIBNQY22QI/graph.json","fetch_events":"https://pith.science/api/pith-number/DEP2WS5QZP5J5BCIYIBNQY22QI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI/action/storage_attestation","attest_author":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI/action/author_attestation","sign_citation":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI/action/citation_signature","submit_replication":"https://pith.science/pith/DEP2WS5QZP5J5BCIYIBNQY22QI/action/replication_record"}},"created_at":"2026-05-20T02:05:43.090455+00:00","updated_at":"2026-05-20T02:05:43.090455+00:00"}