{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:PIC7JTFAWUJQNOUTTPUZBLZMYT","short_pith_number":"pith:PIC7JTFA","schema_version":"1.0","canonical_sha256":"7a05f4cca0b51306ba939be990af2cc4c05baca771fd2319e00bc70735901ba0","source":{"kind":"arxiv","id":"2106.02280","version":1},"attestation_state":"computed","paper":{"title":"Human-Adversarial Visual Question Answering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Amanpreet Singh, Devi Parikh, Douwe Kiela, Jose Alberto Lopez Magana, Sasha Sheng, Vedanuj Goswami, Wojciech Galuba","submitted_at":"2021-06-04T06:25:32Z","abstract_excerpt":"Performance on the most commonly used Visual Question Answering dataset (VQA v2) is starting to approach human accuracy. However, in interacting with state-of-the-art VQA models, it is clear that the problem is far from being solved. In order to stress test VQA models, we benchmark them against human-adversarial examples. Human subjects interact with a state-of-the-art VQA model, and for each image in the dataset, attempt to find a question where the model's predicted answer is incorrect. We find that a wide range of state-of-the-art models perform poorly when evaluated on these examples. We c"},"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":"2106.02280","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2021-06-04T06:25:32Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2832bd1cb2a27008e32e674680c9b3903acf8aef0f04bd001cc70a20e118eb2a","abstract_canon_sha256":"6718b5d3bf7633d9609d61cd016a0d6aea142fde31cb21ba6c88a3fb090655e6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:46:17.674922Z","signature_b64":"hDwOcG738KXruWBnElsGNRaqrvp87HqDSF85+h4fhdCr+NPGJGpr6v+zvxydvSOG54/XAU8ygzOb/TWkbthoCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a05f4cca0b51306ba939be990af2cc4c05baca771fd2319e00bc70735901ba0","last_reissued_at":"2026-07-05T02:46:17.674497Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:46:17.674497Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Human-Adversarial Visual Question Answering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Amanpreet Singh, Devi Parikh, Douwe Kiela, Jose Alberto Lopez Magana, Sasha Sheng, Vedanuj Goswami, Wojciech Galuba","submitted_at":"2021-06-04T06:25:32Z","abstract_excerpt":"Performance on the most commonly used Visual Question Answering dataset (VQA v2) is starting to approach human accuracy. However, in interacting with state-of-the-art VQA models, it is clear that the problem is far from being solved. In order to stress test VQA models, we benchmark them against human-adversarial examples. Human subjects interact with a state-of-the-art VQA model, and for each image in the dataset, attempt to find a question where the model's predicted answer is incorrect. We find that a wide range of state-of-the-art models perform poorly when evaluated on these examples. We c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.02280","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/2106.02280/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":"2106.02280","created_at":"2026-07-05T02:46:17.674558+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.02280v1","created_at":"2026-07-05T02:46:17.674558+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.02280","created_at":"2026-07-05T02:46:17.674558+00:00"},{"alias_kind":"pith_short_12","alias_value":"PIC7JTFAWUJQ","created_at":"2026-07-05T02:46:17.674558+00:00"},{"alias_kind":"pith_short_16","alias_value":"PIC7JTFAWUJQNOUT","created_at":"2026-07-05T02:46:17.674558+00:00"},{"alias_kind":"pith_short_8","alias_value":"PIC7JTFA","created_at":"2026-07-05T02:46:17.674558+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.25863","citing_title":"MCMit: Mid-Circuit Measurement Error Mitigation","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25855","citing_title":"SIEVES: Selective Prediction Generalizes through Visual Evidence Scoring","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25855","citing_title":"SIEVES: Selective Prediction Generalizes through Visual Evidence Scoring","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT","json":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT.json","graph_json":"https://pith.science/api/pith-number/PIC7JTFAWUJQNOUTTPUZBLZMYT/graph.json","events_json":"https://pith.science/api/pith-number/PIC7JTFAWUJQNOUTTPUZBLZMYT/events.json","paper":"https://pith.science/paper/PIC7JTFA"},"agent_actions":{"view_html":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT","download_json":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT.json","view_paper":"https://pith.science/paper/PIC7JTFA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.02280&json=true","fetch_graph":"https://pith.science/api/pith-number/PIC7JTFAWUJQNOUTTPUZBLZMYT/graph.json","fetch_events":"https://pith.science/api/pith-number/PIC7JTFAWUJQNOUTTPUZBLZMYT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT/action/storage_attestation","attest_author":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT/action/author_attestation","sign_citation":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT/action/citation_signature","submit_replication":"https://pith.science/pith/PIC7JTFAWUJQNOUTTPUZBLZMYT/action/replication_record"}},"created_at":"2026-07-05T02:46:17.674558+00:00","updated_at":"2026-07-05T02:46:17.674558+00:00"}