{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:O2T4BEDXJ2BKFT2H4PDBHYSTYI","short_pith_number":"pith:O2T4BEDX","schema_version":"1.0","canonical_sha256":"76a7c090774e82a2cf47e3c613e253c2178648fd662f495f8ae61bbd15f17e73","source":{"kind":"arxiv","id":"2607.14943","version":1},"attestation_state":"computed","paper":{"title":"Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.SY","eess.SY","math.OC"],"primary_cat":"cs.RO","authors_text":"Alice Chan, Glen Chou, Jihoon Hong, Julian Skifstad, Qiyue Dai","submitted_at":"2026-07-16T12:52:49Z","abstract_excerpt":"World Action Models (WAMs) enable semantically- and physically-informed control but are brittle under distribution shift. In this work, we use mechanistic interpretability to study how robustness-relevant perturbations are represented in WAM activation space. Comparing activations across successful and unsuccessful rollouts, we find some WAM architectures exhibit low-dimensional linear separability for robustness-critical features, while others do not. This motivates the use of contrastive activation directions for training-free WAM steering. We also show that local linearity in WAM activation"},"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":"2607.14943","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-07-16T12:52:49Z","cross_cats_sorted":["cs.AI","cs.LG","cs.SY","eess.SY","math.OC"],"title_canon_sha256":"32f50d26cc9e1085af293fd33fc40e868e0d3aaeafbac0d99a22c2a5b6944091","abstract_canon_sha256":"ba06fccc23feda0a38009b74a51858855c0d9686b18324da4b4a5f5fbbf17428"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T01:22:00.362136Z","signature_b64":"8Wyx60TMza23RbGL+HKZ48zB+fmmYWq6ezR74xdUNVeNJEhRrGOeEF3863j6eD2W3eCKf86fqOb/OcTUGhfwAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"76a7c090774e82a2cf47e3c613e253c2178648fd662f495f8ae61bbd15f17e73","last_reissued_at":"2026-07-17T01:22:00.361244Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T01:22:00.361244Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.SY","eess.SY","math.OC"],"primary_cat":"cs.RO","authors_text":"Alice Chan, Glen Chou, Jihoon Hong, Julian Skifstad, Qiyue Dai","submitted_at":"2026-07-16T12:52:49Z","abstract_excerpt":"World Action Models (WAMs) enable semantically- and physically-informed control but are brittle under distribution shift. In this work, we use mechanistic interpretability to study how robustness-relevant perturbations are represented in WAM activation space. Comparing activations across successful and unsuccessful rollouts, we find some WAM architectures exhibit low-dimensional linear separability for robustness-critical features, while others do not. This motivates the use of contrastive activation directions for training-free WAM steering. We also show that local linearity in WAM activation"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14943","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.14943/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":"2607.14943","created_at":"2026-07-17T01:22:00.361703+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14943v1","created_at":"2026-07-17T01:22:00.361703+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14943","created_at":"2026-07-17T01:22:00.361703+00:00"},{"alias_kind":"pith_short_12","alias_value":"O2T4BEDXJ2BK","created_at":"2026-07-17T01:22:00.361703+00:00"},{"alias_kind":"pith_short_16","alias_value":"O2T4BEDXJ2BKFT2H","created_at":"2026-07-17T01:22:00.361703+00:00"},{"alias_kind":"pith_short_8","alias_value":"O2T4BEDX","created_at":"2026-07-17T01:22:00.361703+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/O2T4BEDXJ2BKFT2H4PDBHYSTYI","json":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI.json","graph_json":"https://pith.science/api/pith-number/O2T4BEDXJ2BKFT2H4PDBHYSTYI/graph.json","events_json":"https://pith.science/api/pith-number/O2T4BEDXJ2BKFT2H4PDBHYSTYI/events.json","paper":"https://pith.science/paper/O2T4BEDX"},"agent_actions":{"view_html":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI","download_json":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI.json","view_paper":"https://pith.science/paper/O2T4BEDX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14943&json=true","fetch_graph":"https://pith.science/api/pith-number/O2T4BEDXJ2BKFT2H4PDBHYSTYI/graph.json","fetch_events":"https://pith.science/api/pith-number/O2T4BEDXJ2BKFT2H4PDBHYSTYI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI/action/storage_attestation","attest_author":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI/action/author_attestation","sign_citation":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI/action/citation_signature","submit_replication":"https://pith.science/pith/O2T4BEDXJ2BKFT2H4PDBHYSTYI/action/replication_record"}},"created_at":"2026-07-17T01:22:00.361703+00:00","updated_at":"2026-07-17T01:22:00.361703+00:00"}