{"as_of":"2026-08-08T20:01:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:e9201ff2ff114d99824b25cf947bc1e37adba92d39d2b2c0075136deb7891b29","coverage":[{"denominator":20,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":20,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-04T16:30:29.434614Z","state":"measured"},{"denominator":20,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":20,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-08T06:32:00.761636+00:00","state":"measured"},{"denominator":0,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":0,"source":"paper_references, paper_reference_links","source_observed_at":null,"state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"cited_works","source_observed_at":null,"state":"measured"}],"external_citation_measurements":[],"inbound":[],"links":{"evidence":"/evidence","html":"/paper/2509.13608/citation-record","integrity":"/paper/2509.13608/integrity","json":"/paper/2509.13608/citation-record.json","paper":"/paper/2509.13608"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.327387Z","title":"GPT-4o mini: advancing cost-efficient intelligence","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.327387Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:29c191efdd8b0f8d081c49a3c722da859065bfc12a2f49683961b154a9a9c108","observation_id":"1000e86f-98a5-4714-9d12-ac504861e9dc","resolution":{"observed_at":"2026-08-04T16:30:28.327387Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.378517Z","title":"ChatGPT Statistics 2025","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.378517Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:4a2a9c27337f216f847d9b4dd0b22a171a6ecf3583306288fee1d8659fc4031c","observation_id":"74634001-9b30-416c-b525-439e772a006a","resolution":{"observed_at":"2026-08-04T16:30:28.378517Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.427662Z","title":"The hateful memes challenge: Detecting hate speech in mul- timodal memes","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.427662Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:1bc7fb08776e4015cfb5a6c169550e873a48db3f0cdb499ddf7527e75b1f003a","observation_id":"39719fdb-deff-4cba-b0b9-78b83161c184","resolution":{"observed_at":"2026-08-04T16:30:28.427662Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.511564Z","title":"Cooperative in- verse reinforcement learning.Advances in neural information processing systems, 29, 2016","venue":null,"work_id":null,"year":2016},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.511564Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:f8f29c4d5a24f0028b1991ff5ad2ca0eb5e093112f16ca9bef32218a77497a81","observation_id":"3e9bfc75-46d5-40c5-b377-9e8566541b5f","resolution":{"observed_at":"2026-08-04T16:30:28.511564Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.570854Z","title":"Viking, 2019","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.570854Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:66e32290622ab0a3f10a951063b1206f7f034c09268a4e8680978b2047684180","observation_id":"89eeb79d-e1ac-45fe-a168-562b51cb1e58","resolution":{"observed_at":"2026-08-04T16:30:28.570854Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1810.04805","last_updated":"2019-05-24T20:37:26Z","snapshot_observed_at":"2026-07-30T09:12:38.100527Z","submitted_at":"2018-10-11T00:50:01Z","title":"BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1810.04805","snapshot_observed_at":"2026-08-04T16:30:28.629504Z","title":"BERT: Pre- training of Deep Bidirectional Transformers for Language Understanding.arXiv preprint arXiv:1810.04805, 2018","venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.629504Z"},"links":{"cited_paper":"/paper/1810.04805","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:a0a2b3f8ed5d20fae824f57f5b9fce7096a28c7b0de37fccfcf2c41eebd946d0","observation_id":"6250ec90-87ae-4f03-91a9-c28a2c735827","resolution":{"observed_at":"2026-08-04T16:30:28.629504Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.691674Z","title":"Hate-CLIPper: Multimodal Hateful Meme Classification based on Cross-modal Interaction of CLIP Features","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.691674Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:d46c35c8497665c15e30dd88ba30380741e5745ae798e1788ce60ed309b38d12","observation_id":"59801848-7a23-4750-840d-8d662980be26","resolution":{"observed_at":"2026-08-04T16:30:28.691674Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.740712Z","title":"Singh, and V","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.740712Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:0cd401375c682146d28b84d4834ec8a56a65722878f0b79ccbef09917195e6bd","observation_id":"e09ab6d1-fbea-4974-bb98-1d679ce53eff","resolution":{"observed_at":"2026-08-04T16:30:28.740712Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.803755Z","title":"Multimodal detection of hateful memes by applying a vision-language pre-training model.PLoS one, 16(9):e0257302, 2021","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.803755Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:aee2796bc202a9882c797071416d5c7e617b9efe18e5cc7a3f411e9e6fdf8ad1","observation_id":"d027d822-aaae-4b4f-90a2-0caecebd5e1c","resolution":{"observed_at":"2026-08-04T16:30:28.803755Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.835192Z","title":"Multimodal hate speech detection based on multi-task learning","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.835192Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:26444ceb6beae6da0c0ffd5c202a8ec4c2336cbcb2d83ebbddad7f8ec1217fa3","observation_id":"6b1ac153-4b87-4865-9dfc-307f2ba5b179","resolution":{"observed_at":"2026-08-04T16:30:28.835192Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:28.893257Z","title":"Demystifying Hateful Content: Leveraging Large Multimodal Models for Hateful Meme Detection with Explainable Decisions","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.893257Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:8f537d644401f5c192d8d9273cc42b3dbeeacfae11df5e2c0c2540390930c8ef","observation_id":"4767ef8c-c91b-4531-a401-6adbfcbdb9b6","resolution":{"observed_at":"2026-08-04T16:30:28.893257Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2402.13061","last_updated":"2024-02-20T14:56:28Z","snapshot_observed_at":"2026-07-06T17:32:52.500465Z","submitted_at":"2024-02-20T14:56:28Z","title":"Toward Fairness via Maximum Mean Discrepancy Regularization on Logits Space","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2402.13061","snapshot_observed_at":"2026-08-04T16:30:28.962552Z","title":"Improved Fine-Tuning of Large Multimodal Models for Hateful Meme De- tection.arXiv preprint arXiv:2402.13061, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:28.962552Z"},"links":{"cited_paper":"/paper/2402.13061","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:26f8967540f88cc68b02a03aedb10adf8b8d433adc6eea644f04d8339405acd5","observation_id":"72a80e90-7474-487a-9275-91f58e0d4e25","resolution":{"observed_at":"2026-08-04T16:30:28.962552Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2406.18919","last_updated":"2024-06-27T06:22:02Z","snapshot_observed_at":"2026-07-06T18:37:47.471616Z","submitted_at":"2024-06-27T06:22:02Z","title":"Classification of Carotid Plaque with Jellyfish Sign Through Convolutional and Recurrent Neural Networks Utilizing Plaque Surface Edges","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2406.18919","snapshot_observed_at":"2026-08-04T16:30:29.002464Z","title":"MemeMind: A Large-Scale Multimodal Dataset with Chain-of-Thought Reasoning for Harmful Meme Detection.arXiv preprint arXiv:2406.18919, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.002464Z"},"links":{"cited_paper":"/paper/2406.18919","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:a294bb22298f455d3a388748d8d52cfa9f4f6ad4b4d0d94bf63cf2d7a5b11c44","observation_id":"2d31c79a-d3c7-462f-964f-b89a06b4b6b0","resolution":{"observed_at":"2026-08-04T16:30:29.002464Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:29.084744Z","title":"AI ”safety” vs ”control” vs ”alignment”","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.084744Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:840a0d0cab72155e39095b34b0b621dda07e318c84d74bd040938784b621f0ad","observation_id":"569bbb60-5c61-4576-ae5d-01e637352e70","resolution":{"observed_at":"2026-08-04T16:30:29.084744Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:29.134117Z","title":"How AI is shaping the next generation of brand safety and suitabil- ity","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.134117Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:98fc3b431dde7ba70399650e2f1a7339298d423f3bd43c3e8ebb28bc8e62e27b","observation_id":"7193b448-e006-4a89-8b57-be90ae8a648e","resolution":{"observed_at":"2026-08-04T16:30:29.134117Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2409.13592","last_updated":"2024-09-20T15:45:29Z","snapshot_observed_at":"2026-08-08T07:01:19.029114Z","submitted_at":"2024-09-20T15:45:29Z","title":"YesBut: A High-Quality Annotated Multimodal Dataset for evaluating Satire Comprehension capability of Vision-Language Models","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2409.13592","snapshot_observed_at":"2026-08-04T16:30:29.203246Z","title":"I’m sure you’re a real scholar yourself: Exploring Ironic Content Gen- eration by Large Language Models.arXiv preprint arXiv:2409.13592, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.203246Z"},"links":{"cited_paper":"/paper/2409.13592","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:9d549c9251e49e3a9d008e6ec7a52d936617991f135e6d222cdd6a3acb788a2e","observation_id":"4e3f0002-9a6e-4dd9-bfef-8b569985bc22","resolution":{"observed_at":"2026-08-04T16:30:29.203246Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:29.267971Z","title":"SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Example","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.267971Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:4ab42bdcab8ea2e1b2f5c1a9554d359e9db4de15b2de19389146dcadd800dd76","observation_id":"7b2d6dea-a34c-4679-af48-f1a3f2b8f599","resolution":{"observed_at":"2026-08-04T16:30:29.267971Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2403.11001","last_updated":"2024-10-09T17:44:14Z","snapshot_observed_at":"2026-07-06T17:45:43.726336Z","submitted_at":"2024-03-16T19:11:57Z","title":"Topologically Faithful Multi-class Segmentation in Medical Images","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2403.11001","snapshot_observed_at":"2026-08-04T16:30:29.331818Z","title":"ModSCAN: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities.arXiv preprint arXiv:2403.11001, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.331818Z"},"links":{"cited_paper":"/paper/2403.11001","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:32025e3623da8fba956e60297f6a882e335614598ab807a684873f1764c6f5d1","observation_id":"b8ba8047-e7dc-4314-9691-3abd4de223af","resolution":{"observed_at":"2026-08-04T16:30:29.331818Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2402.00357","last_updated":"2024-06-20T15:06:10Z","snapshot_observed_at":"2026-07-06T17:23:31.057358Z","submitted_at":"2024-02-01T05:57:10Z","title":"Safety of Multimodal Large Language Models on Images and Texts","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2402.00357","snapshot_observed_at":"2026-08-04T16:30:29.384240Z","title":"Safety of Multimodal Large Language Models on Images and Text.arXiv preprint arXiv:2402.00357, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.384240Z"},"links":{"cited_paper":"/paper/2402.00357","citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:aee9d01cb5ec5ad45adfe3b5cf1f6fd2be3140512a84fcc0b2574f927a90b671","observation_id":"3894a040-3054-4e18-a724-87ac4b6121f9","resolution":{"observed_at":"2026-08-04T16:30:29.384240Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T16:30:29.434614Z","title":"Safety of Multimodal Large Language Models on Images and Text","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection","version":2},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-04T16:30:29.434614Z"},"links":{"citing_paper":"/paper/2509.13608"},"observation_digest":"sha256:ebe125bf7e844618d19ce5bb2d35165a3f910bfc94dfbc5d3420bb45cd5298c5","observation_id":"305ba678-17fb-4ee7-891c-e7cc8f01e5e3","resolution":{"observed_at":"2026-08-04T16:30:29.434614Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2509.13608","last_updated":"2026-05-23T00:20:32Z","latest_version":2,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-04T16:30:27.868237Z","submitted_at":"2025-09-17T00:46:42Z","title":"Is GPT-4o mini Blinded by its Own Safety Filters? Exposing the Multimodal-to-Unimodal Bottleneck in Hate Speech Detection"},"reference_resolution":{"displayed":20,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":20,"verified_exact":0,"verified_fuzzy":0},"total_outbound_references":20},"refusal":"A citation records a reference. It does not transfer a finding from one paper to another.","schema":"pith.paper-citation-record.v1","standing_sources":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"thesis":"As of 8 August 2026, this Paper Citation Record lists 20 of 20 outbound references and 0 inbound Pith citation observations for arXiv:2509.13608."}