{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:2M2CIAKOLYNB5DPKDRVO2ISJFW","short_pith_number":"pith:2M2CIAKO","canonical_record":{"source":{"id":"2302.00070","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-31T20:09:33Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"392d2ec66c89c508196cab32f8a29a46a816871db0c3ef5fa1ad255acdb03a5a","abstract_canon_sha256":"63eba94affaa96d902ab444001bad17b36de1bd3e315031e53674978a3d74ea5"},"schema_version":"1.0"},"canonical_sha256":"d33424014e5e1a1e8dea1c6aed22492d9f7adccc9c039fb6ebdaf9bdd3517cb1","source":{"kind":"arxiv","id":"2302.00070","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2302.00070","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"arxiv_version","alias_value":"2302.00070v2","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.00070","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_12","alias_value":"2M2CIAKOLYNB","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_16","alias_value":"2M2CIAKOLYNB5DPK","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_8","alias_value":"2M2CIAKO","created_at":"2026-07-05T06:09:47Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:2M2CIAKOLYNB5DPKDRVO2ISJFW","target":"record","payload":{"canonical_record":{"source":{"id":"2302.00070","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-31T20:09:33Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"392d2ec66c89c508196cab32f8a29a46a816871db0c3ef5fa1ad255acdb03a5a","abstract_canon_sha256":"63eba94affaa96d902ab444001bad17b36de1bd3e315031e53674978a3d74ea5"},"schema_version":"1.0"},"canonical_sha256":"d33424014e5e1a1e8dea1c6aed22492d9f7adccc9c039fb6ebdaf9bdd3517cb1","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:09:47.970644Z","signature_b64":"VECSHVMQn7zcDTpZmNPXRXhtkcicanuQtK/+MaUr9CPMbDaIRRJZPfFNugqGHYgM3LnKYiw4Kf8qL0K8a3SkDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d33424014e5e1a1e8dea1c6aed22492d9f7adccc9c039fb6ebdaf9bdd3517cb1","last_reissued_at":"2026-07-05T06:09:47.970224Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:09:47.970224Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2302.00070","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:09:47Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7IFW6o7wNBdlcKKMKzoOoxL9tQfOxdGpWoU3H9vNyuiUYgSrf/TLdWEZnT4CbMkklJpb6Nb/iniJDJMo+B08Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T16:27:41.567065Z"},"content_sha256":"bbc2178ca5acd3f027fc7c2eff70a7c3c96aae4f491efe03c5f1c26d19d00c81","schema_version":"1.0","event_id":"sha256:bbc2178ca5acd3f027fc7c2eff70a7c3c96aae4f491efe03c5f1c26d19d00c81"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:2M2CIAKOLYNB5DPKDRVO2ISJFW","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Debiasing Vision-Language Models via Biased Prompts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Antonio Torralba, Ching-Yao Chuang, Stefanie Jegelka, Varun Jampani, Yuanzhen Li","submitted_at":"2023-01-31T20:09:33Z","abstract_excerpt":"Machine learning models have been shown to inherit biases from their training datasets. This can be particularly problematic for vision-language foundation models trained on uncurated datasets scraped from the internet. The biases can be amplified and propagated to downstream applications like zero-shot classifiers and text-to-image generative models. In this study, we propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. In particular, we show that debiasing only the text embedding with a calibrated projection mat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.00070","kind":"arxiv","version":2},"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/2302.00070/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:09:47Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"2bl77A6MUbo76R6HJijcx0I55Vvv9i2skx8dKBJ3WSpeh0+q9QUh0GM8e7UTX3BYxdNdIAY2vGpjXWa2VFJqDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T16:27:41.567576Z"},"content_sha256":"3f18d2da17a38e1d40c66e951fd72d699c9bd91d4d01714dc1af21d50e165738","schema_version":"1.0","event_id":"sha256:3f18d2da17a38e1d40c66e951fd72d699c9bd91d4d01714dc1af21d50e165738"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/bundle.json","state_url":"https://pith.science/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-07T16:27:41Z","links":{"resolver":"https://pith.science/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW","bundle":"https://pith.science/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/bundle.json","state":"https://pith.science/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/state.json","well_known_bundle":"https://pith.science/.well-known/pith/2M2CIAKOLYNB5DPKDRVO2ISJFW/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:2M2CIAKOLYNB5DPKDRVO2ISJFW","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"63eba94affaa96d902ab444001bad17b36de1bd3e315031e53674978a3d74ea5","cross_cats_sorted":["cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-31T20:09:33Z","title_canon_sha256":"392d2ec66c89c508196cab32f8a29a46a816871db0c3ef5fa1ad255acdb03a5a"},"schema_version":"1.0","source":{"id":"2302.00070","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2302.00070","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"arxiv_version","alias_value":"2302.00070v2","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.00070","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_12","alias_value":"2M2CIAKOLYNB","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_16","alias_value":"2M2CIAKOLYNB5DPK","created_at":"2026-07-05T06:09:47Z"},{"alias_kind":"pith_short_8","alias_value":"2M2CIAKO","created_at":"2026-07-05T06:09:47Z"}],"graph_snapshots":[{"event_id":"sha256:3f18d2da17a38e1d40c66e951fd72d699c9bd91d4d01714dc1af21d50e165738","target":"graph","created_at":"2026-07-05T06:09:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2302.00070/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Machine learning models have been shown to inherit biases from their training datasets. This can be particularly problematic for vision-language foundation models trained on uncurated datasets scraped from the internet. The biases can be amplified and propagated to downstream applications like zero-shot classifiers and text-to-image generative models. In this study, we propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. In particular, we show that debiasing only the text embedding with a calibrated projection mat","authors_text":"Antonio Torralba, Ching-Yao Chuang, Stefanie Jegelka, Varun Jampani, Yuanzhen Li","cross_cats":["cs.CV"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-31T20:09:33Z","title":"Debiasing Vision-Language Models via Biased Prompts"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.00070","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:bbc2178ca5acd3f027fc7c2eff70a7c3c96aae4f491efe03c5f1c26d19d00c81","target":"record","created_at":"2026-07-05T06:09:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"63eba94affaa96d902ab444001bad17b36de1bd3e315031e53674978a3d74ea5","cross_cats_sorted":["cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-31T20:09:33Z","title_canon_sha256":"392d2ec66c89c508196cab32f8a29a46a816871db0c3ef5fa1ad255acdb03a5a"},"schema_version":"1.0","source":{"id":"2302.00070","kind":"arxiv","version":2}},"canonical_sha256":"d33424014e5e1a1e8dea1c6aed22492d9f7adccc9c039fb6ebdaf9bdd3517cb1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d33424014e5e1a1e8dea1c6aed22492d9f7adccc9c039fb6ebdaf9bdd3517cb1","first_computed_at":"2026-07-05T06:09:47.970224Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T06:09:47.970224Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VECSHVMQn7zcDTpZmNPXRXhtkcicanuQtK/+MaUr9CPMbDaIRRJZPfFNugqGHYgM3LnKYiw4Kf8qL0K8a3SkDA==","signature_status":"signed_v1","signed_at":"2026-07-05T06:09:47.970644Z","signed_message":"canonical_sha256_bytes"},"source_id":"2302.00070","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bbc2178ca5acd3f027fc7c2eff70a7c3c96aae4f491efe03c5f1c26d19d00c81","sha256:3f18d2da17a38e1d40c66e951fd72d699c9bd91d4d01714dc1af21d50e165738"],"state_sha256":"9214f15dac53cd4129d7bda27496db0bce57dcbdec1d4e903b10635ddc7b903e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BfyhgL5wHPA+9gNrXWBwmF2IFnx3CeF4RkIt5NGDHQsiRr//fwvIv3jrM4ewl7qIAzJE4ncuNgtxTq2FdtHXAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T16:27:41.572656Z","bundle_sha256":"32f600c7b8cee208df5b343a38b5522c2915263ef7276650b4d14bb4cae7c252"}}