{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AAJ5LLTZBPTBNTJDWNPOQDPM3M","short_pith_number":"pith:AAJ5LLTZ","schema_version":"1.0","canonical_sha256":"0013d5ae790be616cd23b35ee80decdb1a6047d94dc0eb413b6e87e45296668a","source":{"kind":"arxiv","id":"2607.13429","version":1},"attestation_state":"computed","paper":{"title":"Generalizable VLA Finetuning via Representation Anchoring and Language-Action Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Alex Baratian, Chahit Jain, Dwip Dalal, Heng Ji, Hyeonjeong Ha, Jeonghwan Kim, Shivansh Patel, Svetlana Lazebnik, Unnat Jain, Utkarsh Mishra","submitted_at":"2026-07-15T04:13:54Z","abstract_excerpt":"Finetuning a pretrained vision-language model (VLM) on robot demonstrations via behavior cloning (BC) has become the standard recipe for vision-language-action (VLA) policies. However, BC finetuning progressively overwrites the pretrained representations that support visual and semantic generalization. Co-training on web image-text data, a common remedy, does not prevent this; it applies language and action losses to separate observations, leaving VLAs with language-action misalignment that standard manipulation benchmarks do not expose. We propose Anchor-Align, which augments BC with two obje"},"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":"2607.13429","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-07-15T04:13:54Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"f46f36d48be6c255094abbd306bc1b5602e1e8952e92c590fee9358e31fde849","abstract_canon_sha256":"950860d3567ac25c9b87688990b554819feb133cdc294861fb939ad94ec6f609"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-16T00:22:17.796826Z","signature_b64":"JTdEwe8zihqVJRd1DIWlcwerkTfMZGFmy2c1sGb6/9h8FHY20IxqsI+4B3rzjOft3PLLabbZTC/8PFhCsEnKBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0013d5ae790be616cd23b35ee80decdb1a6047d94dc0eb413b6e87e45296668a","last_reissued_at":"2026-07-16T00:22:17.795817Z","signature_status":"signed_v1","first_computed_at":"2026-07-16T00:22:17.795817Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Generalizable VLA Finetuning via Representation Anchoring and Language-Action Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Alex Baratian, Chahit Jain, Dwip Dalal, Heng Ji, Hyeonjeong Ha, Jeonghwan Kim, Shivansh Patel, Svetlana Lazebnik, Unnat Jain, Utkarsh Mishra","submitted_at":"2026-07-15T04:13:54Z","abstract_excerpt":"Finetuning a pretrained vision-language model (VLM) on robot demonstrations via behavior cloning (BC) has become the standard recipe for vision-language-action (VLA) policies. However, BC finetuning progressively overwrites the pretrained representations that support visual and semantic generalization. Co-training on web image-text data, a common remedy, does not prevent this; it applies language and action losses to separate observations, leaving VLAs with language-action misalignment that standard manipulation benchmarks do not expose. We propose Anchor-Align, which augments BC with two obje"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.13429","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/2607.13429/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":"2607.13429","created_at":"2026-07-16T00:22:17.796303+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.13429v1","created_at":"2026-07-16T00:22:17.796303+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.13429","created_at":"2026-07-16T00:22:17.796303+00:00"},{"alias_kind":"pith_short_12","alias_value":"AAJ5LLTZBPTB","created_at":"2026-07-16T00:22:17.796303+00:00"},{"alias_kind":"pith_short_16","alias_value":"AAJ5LLTZBPTBNTJD","created_at":"2026-07-16T00:22:17.796303+00:00"},{"alias_kind":"pith_short_8","alias_value":"AAJ5LLTZ","created_at":"2026-07-16T00:22:17.796303+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/AAJ5LLTZBPTBNTJDWNPOQDPM3M","json":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M.json","graph_json":"https://pith.science/api/pith-number/AAJ5LLTZBPTBNTJDWNPOQDPM3M/graph.json","events_json":"https://pith.science/api/pith-number/AAJ5LLTZBPTBNTJDWNPOQDPM3M/events.json","paper":"https://pith.science/paper/AAJ5LLTZ"},"agent_actions":{"view_html":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M","download_json":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M.json","view_paper":"https://pith.science/paper/AAJ5LLTZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.13429&json=true","fetch_graph":"https://pith.science/api/pith-number/AAJ5LLTZBPTBNTJDWNPOQDPM3M/graph.json","fetch_events":"https://pith.science/api/pith-number/AAJ5LLTZBPTBNTJDWNPOQDPM3M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M/action/storage_attestation","attest_author":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M/action/author_attestation","sign_citation":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M/action/citation_signature","submit_replication":"https://pith.science/pith/AAJ5LLTZBPTBNTJDWNPOQDPM3M/action/replication_record"}},"created_at":"2026-07-16T00:22:17.796303+00:00","updated_at":"2026-07-16T00:22:17.796303+00:00"}