{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:LGBI7AJKRTDU6BSXWKQTE57BZ4","short_pith_number":"pith:LGBI7AJK","schema_version":"1.0","canonical_sha256":"59828f812a8cc74f0657b2a13277e1cf0a5bd4ded01d8b11f3cd2bc59e759576","source":{"kind":"arxiv","id":"2605.21642","version":1},"attestation_state":"computed","paper":{"title":"Ablate-to-Validate: Are Vision-Language Models Really Using Continuous Thought Tokens?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Mahtab Bigverdi, Ranjay Krishna, Tianyi Zhang","submitted_at":"2026-05-20T18:55:16Z","abstract_excerpt":"Vision-language models (VLMs) are increasingly augmented with continuous or latent non-textual tokens intended to support \"visual thinking.\" Despite improved task accuracy, this alone does not show that models actually use these tokens for reasoning -- gains may arise from confounds such as added context length, special-token anchoring, or training-time regularization. We formalize a diagnostic principle, Ablate-to-Validate, for testing whether latent-token content is genuinely utilized, and instantiate it as the Token Replacement Test (TRT), a standardized suite of content-replacement ablatio"},"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":"2605.21642","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-20T18:55:16Z","cross_cats_sorted":[],"title_canon_sha256":"548240e39f8927c883d08f8244e18cb9c0e8b8109ea479cf7e3d6c6869e50e49","abstract_canon_sha256":"efcc5917da086cf3257d23ea65beb4770fe4a9de1a2326855d49a228e2e4f772"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:26.569667Z","signature_b64":"mf1R2r+bWZc6qJV1z/p+/FXzPhvZ0l5l3XH92q2MfV7sZK1dRw8HKkSuxyAqTE2EDGtr/VE5c21Q79mJStTACA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"59828f812a8cc74f0657b2a13277e1cf0a5bd4ded01d8b11f3cd2bc59e759576","last_reissued_at":"2026-05-22T01:03:26.569172Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:26.569172Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Ablate-to-Validate: Are Vision-Language Models Really Using Continuous Thought Tokens?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Mahtab Bigverdi, Ranjay Krishna, Tianyi Zhang","submitted_at":"2026-05-20T18:55:16Z","abstract_excerpt":"Vision-language models (VLMs) are increasingly augmented with continuous or latent non-textual tokens intended to support \"visual thinking.\" Despite improved task accuracy, this alone does not show that models actually use these tokens for reasoning -- gains may arise from confounds such as added context length, special-token anchoring, or training-time regularization. We formalize a diagnostic principle, Ablate-to-Validate, for testing whether latent-token content is genuinely utilized, and instantiate it as the Token Replacement Test (TRT), a standardized suite of content-replacement ablatio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21642","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/2605.21642/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":"2605.21642","created_at":"2026-05-22T01:03:26.569249+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21642v1","created_at":"2026-05-22T01:03:26.569249+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21642","created_at":"2026-05-22T01:03:26.569249+00:00"},{"alias_kind":"pith_short_12","alias_value":"LGBI7AJKRTDU","created_at":"2026-05-22T01:03:26.569249+00:00"},{"alias_kind":"pith_short_16","alias_value":"LGBI7AJKRTDU6BSX","created_at":"2026-05-22T01:03:26.569249+00:00"},{"alias_kind":"pith_short_8","alias_value":"LGBI7AJK","created_at":"2026-05-22T01:03:26.569249+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/LGBI7AJKRTDU6BSXWKQTE57BZ4","json":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4.json","graph_json":"https://pith.science/api/pith-number/LGBI7AJKRTDU6BSXWKQTE57BZ4/graph.json","events_json":"https://pith.science/api/pith-number/LGBI7AJKRTDU6BSXWKQTE57BZ4/events.json","paper":"https://pith.science/paper/LGBI7AJK"},"agent_actions":{"view_html":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4","download_json":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4.json","view_paper":"https://pith.science/paper/LGBI7AJK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21642&json=true","fetch_graph":"https://pith.science/api/pith-number/LGBI7AJKRTDU6BSXWKQTE57BZ4/graph.json","fetch_events":"https://pith.science/api/pith-number/LGBI7AJKRTDU6BSXWKQTE57BZ4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4/action/storage_attestation","attest_author":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4/action/author_attestation","sign_citation":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4/action/citation_signature","submit_replication":"https://pith.science/pith/LGBI7AJKRTDU6BSXWKQTE57BZ4/action/replication_record"}},"created_at":"2026-05-22T01:03:26.569249+00:00","updated_at":"2026-05-22T01:03:26.569249+00:00"}