{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7HAXDK4ZUOACMUPRUKJGBNGJUK","short_pith_number":"pith:7HAXDK4Z","schema_version":"1.0","canonical_sha256":"f9c171ab99a3802651f1a29260b4c9a296ba7d65dfdb76a85511530fb7b355f1","source":{"kind":"arxiv","id":"2606.08777","version":1},"attestation_state":"computed","paper":{"title":"How Many Counterfactuals Does It Take? Probing VLM Hallucinations Through Circuits and Causal Effects","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Abhivansh Gupta, Advika Sinha, Akshat Tomar, Shreyansh Modi, Simardeep Singh","submitted_at":"2026-06-07T18:45:29Z","abstract_excerpt":"Visual Language Models (VLMs) are known to produce hallucinated predictions that are not grounded in visual evidence, yet existing approaches lack a principled understanding of how robust such predictions are under counterfactual perturbations. In this work, we study the sample complexity of counterfactual robustness for hallucinated outputs in VLMs. We define a causal influence metric based on log-probability differences between factual, counterfactual, and activation-patched runs, and use it to characterize the stability of hallucinated predictions. By leveraging circuit discovery techniques"},"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":"2606.08777","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-07T18:45:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7a7fa9bbf139d13efd45473c298d648a7d96e4f5c870315667ca8453a794d7bb","abstract_canon_sha256":"c3a9660f8bfbb0fe8be4a6890cf48593479262260b30cc1bf4b9c3a89846610a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:07:38.871201Z","signature_b64":"NQSjhOvl6/FUttqru9Gl0OPwBD2GsKNxHTCxnEdB8ULGkk4SICq8LwalDYZYDqoB+W5OAH7rgqrDAgFG7Xq/Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f9c171ab99a3802651f1a29260b4c9a296ba7d65dfdb76a85511530fb7b355f1","last_reissued_at":"2026-06-09T02:07:38.870383Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:07:38.870383Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Many Counterfactuals Does It Take? Probing VLM Hallucinations Through Circuits and Causal Effects","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Abhivansh Gupta, Advika Sinha, Akshat Tomar, Shreyansh Modi, Simardeep Singh","submitted_at":"2026-06-07T18:45:29Z","abstract_excerpt":"Visual Language Models (VLMs) are known to produce hallucinated predictions that are not grounded in visual evidence, yet existing approaches lack a principled understanding of how robust such predictions are under counterfactual perturbations. In this work, we study the sample complexity of counterfactual robustness for hallucinated outputs in VLMs. We define a causal influence metric based on log-probability differences between factual, counterfactual, and activation-patched runs, and use it to characterize the stability of hallucinated predictions. By leveraging circuit discovery techniques"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08777","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/2606.08777/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":"2606.08777","created_at":"2026-06-09T02:07:38.870509+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08777v1","created_at":"2026-06-09T02:07:38.870509+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08777","created_at":"2026-06-09T02:07:38.870509+00:00"},{"alias_kind":"pith_short_12","alias_value":"7HAXDK4ZUOAC","created_at":"2026-06-09T02:07:38.870509+00:00"},{"alias_kind":"pith_short_16","alias_value":"7HAXDK4ZUOACMUPR","created_at":"2026-06-09T02:07:38.870509+00:00"},{"alias_kind":"pith_short_8","alias_value":"7HAXDK4Z","created_at":"2026-06-09T02:07:38.870509+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/7HAXDK4ZUOACMUPRUKJGBNGJUK","json":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK.json","graph_json":"https://pith.science/api/pith-number/7HAXDK4ZUOACMUPRUKJGBNGJUK/graph.json","events_json":"https://pith.science/api/pith-number/7HAXDK4ZUOACMUPRUKJGBNGJUK/events.json","paper":"https://pith.science/paper/7HAXDK4Z"},"agent_actions":{"view_html":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK","download_json":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK.json","view_paper":"https://pith.science/paper/7HAXDK4Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08777&json=true","fetch_graph":"https://pith.science/api/pith-number/7HAXDK4ZUOACMUPRUKJGBNGJUK/graph.json","fetch_events":"https://pith.science/api/pith-number/7HAXDK4ZUOACMUPRUKJGBNGJUK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK/action/storage_attestation","attest_author":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK/action/author_attestation","sign_citation":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK/action/citation_signature","submit_replication":"https://pith.science/pith/7HAXDK4ZUOACMUPRUKJGBNGJUK/action/replication_record"}},"created_at":"2026-06-09T02:07:38.870509+00:00","updated_at":"2026-06-09T02:07:38.870509+00:00"}