{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XPATLXYNGQNJA2M7MSX7K5O5GW","short_pith_number":"pith:XPATLXYN","schema_version":"1.0","canonical_sha256":"bbc135df0d341a90699f64aff575dd359e99c3557d9238c7c8950661cf34fc9a","source":{"kind":"arxiv","id":"2509.00328","version":1},"attestation_state":"computed","paper":{"title":"Mechanistic interpretability for steering vision-language-action models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.RO","authors_text":"Bear H\\\"aon, Claire Tomlin, Ian Chuang, Kaylene Stocking","submitted_at":"2025-08-30T03:01:57Z","abstract_excerpt":"Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents that can quickly adapt to new tasks, modalities, and environments. However, methods for interpreting and steering VLAs fall far short of classical robotics pipelines, which are grounded in explicit models of kinematics, dynamics, and control. This lack of mechanistic insight is a central challenge for deploying learned policies in real-world robotics, where robustness and explainability are critical. Motivated by advances in mechanistic interpretability for large language models, we introduce the f"},"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":"2509.00328","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2025-08-30T03:01:57Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"4e54cb6b9d933d752853b3f088fd284bf544bb6647cc99270557afdda9b07d1e","abstract_canon_sha256":"d9f1d4932e3f3f1be4a1985e9046a3fbfb15488011f5ef707e0b09f8174735ba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:02:13.801694Z","signature_b64":"cHkTbneazO/LXljNkMvgRie7s5Zh6Z+T/g1OofUMNJ3V7Xy4jYYg4Oe/8QcrWq1mio6rh9vJUWegclneHyuPDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bbc135df0d341a90699f64aff575dd359e99c3557d9238c7c8950661cf34fc9a","last_reissued_at":"2026-07-05T12:02:13.801177Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:02:13.801177Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mechanistic interpretability for steering vision-language-action models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.RO","authors_text":"Bear H\\\"aon, Claire Tomlin, Ian Chuang, Kaylene Stocking","submitted_at":"2025-08-30T03:01:57Z","abstract_excerpt":"Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents that can quickly adapt to new tasks, modalities, and environments. However, methods for interpreting and steering VLAs fall far short of classical robotics pipelines, which are grounded in explicit models of kinematics, dynamics, and control. This lack of mechanistic insight is a central challenge for deploying learned policies in real-world robotics, where robustness and explainability are critical. Motivated by advances in mechanistic interpretability for large language models, we introduce the f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.00328","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/2509.00328/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":"2509.00328","created_at":"2026-07-05T12:02:13.801238+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.00328v1","created_at":"2026-07-05T12:02:13.801238+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.00328","created_at":"2026-07-05T12:02:13.801238+00:00"},{"alias_kind":"pith_short_12","alias_value":"XPATLXYNGQNJ","created_at":"2026-07-05T12:02:13.801238+00:00"},{"alias_kind":"pith_short_16","alias_value":"XPATLXYNGQNJA2M7","created_at":"2026-07-05T12:02:13.801238+00:00"},{"alias_kind":"pith_short_8","alias_value":"XPATLXYN","created_at":"2026-07-05T12:02:13.801238+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26588","citing_title":"Inference-Time Robot Behavior Steering through Physically-Aware Reconfiguration of Task-Structure","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12299","citing_title":"Learning What to Say to Your VLA: Mostly Harmless Vision Language Action Model Steering","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10495","citing_title":"Act on What You See: Unlocking Safe Social Navigation in Vision-Language-Action Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29201","citing_title":"Behavior Uncloning: Distilling Mode Redirection into Policy Weights without Inference-Time Steering","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29699","citing_title":"Early Warning Signals for OpenVLA Failure under Visual Distribution Shift","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17204","citing_title":"Event-Grounded Sparse Autoencoders for Vision-Language-Action Policies","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02965","citing_title":"Open-Loop Planning, Closed-Loop Verification: Speculative Verification for VLA","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19018","citing_title":"Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Control","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW","json":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW.json","graph_json":"https://pith.science/api/pith-number/XPATLXYNGQNJA2M7MSX7K5O5GW/graph.json","events_json":"https://pith.science/api/pith-number/XPATLXYNGQNJA2M7MSX7K5O5GW/events.json","paper":"https://pith.science/paper/XPATLXYN"},"agent_actions":{"view_html":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW","download_json":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW.json","view_paper":"https://pith.science/paper/XPATLXYN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.00328&json=true","fetch_graph":"https://pith.science/api/pith-number/XPATLXYNGQNJA2M7MSX7K5O5GW/graph.json","fetch_events":"https://pith.science/api/pith-number/XPATLXYNGQNJA2M7MSX7K5O5GW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW/action/storage_attestation","attest_author":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW/action/author_attestation","sign_citation":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW/action/citation_signature","submit_replication":"https://pith.science/pith/XPATLXYNGQNJA2M7MSX7K5O5GW/action/replication_record"}},"created_at":"2026-07-05T12:02:13.801238+00:00","updated_at":"2026-07-05T12:02:13.801238+00:00"}