{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:LEJDCFCS5A2MME6EQ7OHRMB46N","short_pith_number":"pith:LEJDCFCS","canonical_record":{"source":{"id":"2604.17376","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-19T10:59:17Z","cross_cats_sorted":["cs.AI","cs.LG","eess.IV"],"title_canon_sha256":"46b86415a11ec011f0f1e458950606d6017e8ea8f5419af815ce7e0c8a1b253d","abstract_canon_sha256":"acd47dc827092dd185d3ffc31ceab8b00b0599817e44259923f06958a6a98af5"},"schema_version":"1.0"},"canonical_sha256":"5912311452e834c613c487dc78b03cf37b7de9631b6f590b74a218d2fbac039b","source":{"kind":"arxiv","id":"2604.17376","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17376","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17376v2","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17376","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_12","alias_value":"LEJDCFCS5A2M","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_16","alias_value":"LEJDCFCS5A2MME6E","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_8","alias_value":"LEJDCFCS","created_at":"2026-07-07T03:18:55Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:LEJDCFCS5A2MME6EQ7OHRMB46N","target":"record","payload":{"canonical_record":{"source":{"id":"2604.17376","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-19T10:59:17Z","cross_cats_sorted":["cs.AI","cs.LG","eess.IV"],"title_canon_sha256":"46b86415a11ec011f0f1e458950606d6017e8ea8f5419af815ce7e0c8a1b253d","abstract_canon_sha256":"acd47dc827092dd185d3ffc31ceab8b00b0599817e44259923f06958a6a98af5"},"schema_version":"1.0"},"canonical_sha256":"5912311452e834c613c487dc78b03cf37b7de9631b6f590b74a218d2fbac039b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T03:18:55.918955Z","signature_b64":"hfAgjNYRis5GTIjhXp4HgBKv6gxs318e6+wZ3ZaMTXXpuTtrbAMN+T3Sr/X8j2lbPN0ZHLsGbLo7Tjcj1S9EAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5912311452e834c613c487dc78b03cf37b7de9631b6f590b74a218d2fbac039b","last_reissued_at":"2026-07-07T03:18:55.918443Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T03:18:55.918443Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.17376","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":"2026-07-07T03:18:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"yUUf+CT6/5RCawfv4WYYctRwoTVD4JnriC3M9WJjyNesl5MDBoA9lsE9ifzV0e/Uf/NAHnbwc3hO+wLhCQ+aAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T22:09:23.769395Z"},"content_sha256":"7c84fb99c35c6c4b53878529bc17d6b2c4b743ad70ec96b6767139cfa5e5848b","schema_version":"1.0","event_id":"sha256:7c84fb99c35c6c4b53878529bc17d6b2c4b743ad70ec96b6767139cfa5e5848b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:LEJDCFCS5A2MME6EQ7OHRMB46N","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Towards Generalizable Deepfake Image Detection with Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations.","cross_cats":["cs.AI","cs.LG","eess.IV"],"primary_cat":"cs.CV","authors_text":"Aryan Herur, Deepu Vijayasenan, Hemanth K Mogilipalem, Jayavarapu S Abhinai, Kaliki V Srinanda, M Manvith Prabhu, Vaibhav Santhosh","submitted_at":"2026-04-19T10:59:17Z","abstract_excerpt":"In today's day and age, we face a challenge in detecting deepfake images because of the fast evolution of modern generative models and the poor generalization capability of existing methods. In this paper, we use an ensemble of fine-tuned vision transformers like DINOv2, AIMv2 and OpenCLIP's ViT-L/14 to create generalizable method to detect deepfakes. We use the DF-Wild dataset released as part of the IEEE SP Cup 2025, because it uses a challenging and diverse set of manipulations and generation techniques. We started our experiments with CNN classifiers trained on spatial features. Experiment"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"our ensemble outperforms individual models and strong CNN baselines, achieving an AUC of 96.77% and an Equal Error Rate (EER) of just 9% on the DF-Wild test set, beating the state-of-the-art deepfake detection algorithm Effort by 7.05% and 8% in AUC and EER respectively.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The DF-Wild dataset's mix of manipulations is representative enough that fine-tuning these particular vision transformers will continue to generalize to future unseen deepfake generation techniques.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Ensemble of vision transformers reaches 96.77% AUC and 9% EER on DF-Wild deepfake test set, outperforming the prior Effort baseline by 7% AUC and 8% EER.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"217a4853c23e863d97fd75b5ac60276c68e9cb20baa8882b943d66a6f349f788"},"source":{"id":"2604.17376","kind":"arxiv","version":2},"verdict":{"id":"12634908-5081-4945-b98c-dc4d09751379","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T06:32:15.669893Z","strongest_claim":"our ensemble outperforms individual models and strong CNN baselines, achieving an AUC of 96.77% and an Equal Error Rate (EER) of just 9% on the DF-Wild test set, beating the state-of-the-art deepfake detection algorithm Effort by 7.05% and 8% in AUC and EER respectively.","one_line_summary":"Ensemble of vision transformers reaches 96.77% AUC and 9% EER on DF-Wild deepfake test set, outperforming the prior Effort baseline by 7% AUC and 8% EER.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The DF-Wild dataset's mix of manipulations is representative enough that fine-tuning these particular vision transformers will continue to generalize to future unseen deepfake generation techniques.","pith_extraction_headline":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.17376/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":"12634908-5081-4945-b98c-dc4d09751379"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-07T03:18:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"pG//NTWKbZDfpEvAEF6OZduurIvQwINusub9E/XytNuDlXLIMThv7MPBo3N6IFPk3X+ID3CBW5YZLwmLj8XABA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T22:09:23.770016Z"},"content_sha256":"fd19d3829b652eefc6a702e244e19ea0de80fa7ff784a82609910ddbd83c7417","schema_version":"1.0","event_id":"sha256:fd19d3829b652eefc6a702e244e19ea0de80fa7ff784a82609910ddbd83c7417"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/bundle.json","state_url":"https://pith.science/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/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":"2026-08-04T22:09:23Z","links":{"resolver":"https://pith.science/pith/LEJDCFCS5A2MME6EQ7OHRMB46N","bundle":"https://pith.science/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/bundle.json","state":"https://pith.science/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/state.json","well_known_bundle":"https://pith.science/.well-known/pith/LEJDCFCS5A2MME6EQ7OHRMB46N/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:LEJDCFCS5A2MME6EQ7OHRMB46N","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":"acd47dc827092dd185d3ffc31ceab8b00b0599817e44259923f06958a6a98af5","cross_cats_sorted":["cs.AI","cs.LG","eess.IV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-19T10:59:17Z","title_canon_sha256":"46b86415a11ec011f0f1e458950606d6017e8ea8f5419af815ce7e0c8a1b253d"},"schema_version":"1.0","source":{"id":"2604.17376","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17376","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17376v2","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17376","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_12","alias_value":"LEJDCFCS5A2M","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_16","alias_value":"LEJDCFCS5A2MME6E","created_at":"2026-07-07T03:18:55Z"},{"alias_kind":"pith_short_8","alias_value":"LEJDCFCS","created_at":"2026-07-07T03:18:55Z"}],"graph_snapshots":[{"event_id":"sha256:fd19d3829b652eefc6a702e244e19ea0de80fa7ff784a82609910ddbd83c7417","target":"graph","created_at":"2026-07-07T03:18:55Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"our ensemble outperforms individual models and strong CNN baselines, achieving an AUC of 96.77% and an Equal Error Rate (EER) of just 9% on the DF-Wild test set, beating the state-of-the-art deepfake detection algorithm Effort by 7.05% and 8% in AUC and EER respectively."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The DF-Wild dataset's mix of manipulations is representative enough that fine-tuning these particular vision transformers will continue to generalize to future unseen deepfake generation techniques."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Ensemble of vision transformers reaches 96.77% AUC and 9% EER on DF-Wild deepfake test set, outperforming the prior Effort baseline by 7% AUC and 8% EER."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations."}],"snapshot_sha256":"217a4853c23e863d97fd75b5ac60276c68e9cb20baa8882b943d66a6f349f788"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.17376/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"In today's day and age, we face a challenge in detecting deepfake images because of the fast evolution of modern generative models and the poor generalization capability of existing methods. In this paper, we use an ensemble of fine-tuned vision transformers like DINOv2, AIMv2 and OpenCLIP's ViT-L/14 to create generalizable method to detect deepfakes. We use the DF-Wild dataset released as part of the IEEE SP Cup 2025, because it uses a challenging and diverse set of manipulations and generation techniques. We started our experiments with CNN classifiers trained on spatial features. Experiment","authors_text":"Aryan Herur, Deepu Vijayasenan, Hemanth K Mogilipalem, Jayavarapu S Abhinai, Kaliki V Srinanda, M Manvith Prabhu, Vaibhav Santhosh","cross_cats":["cs.AI","cs.LG","eess.IV"],"headline":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-19T10:59:17Z","title":"Towards Generalizable Deepfake Image Detection with Vision Transformers"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.17376","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T06:32:15.669893Z","id":"12634908-5081-4945-b98c-dc4d09751379","model_set":{"reader":"grok-4.3"},"one_line_summary":"Ensemble of vision transformers reaches 96.77% AUC and 9% EER on DF-Wild deepfake test set, outperforming the prior Effort baseline by 7% AUC and 8% EER.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"An ensemble of fine-tuned vision transformers detects deepfake images with 96.77% AUC on a diverse test set of manipulations.","strongest_claim":"our ensemble outperforms individual models and strong CNN baselines, achieving an AUC of 96.77% and an Equal Error Rate (EER) of just 9% on the DF-Wild test set, beating the state-of-the-art deepfake detection algorithm Effort by 7.05% and 8% in AUC and EER respectively.","weakest_assumption":"The DF-Wild dataset's mix of manipulations is representative enough that fine-tuning these particular vision transformers will continue to generalize to future unseen deepfake generation techniques."}},"verdict_id":"12634908-5081-4945-b98c-dc4d09751379"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:7c84fb99c35c6c4b53878529bc17d6b2c4b743ad70ec96b6767139cfa5e5848b","target":"record","created_at":"2026-07-07T03:18:55Z","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":"acd47dc827092dd185d3ffc31ceab8b00b0599817e44259923f06958a6a98af5","cross_cats_sorted":["cs.AI","cs.LG","eess.IV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-19T10:59:17Z","title_canon_sha256":"46b86415a11ec011f0f1e458950606d6017e8ea8f5419af815ce7e0c8a1b253d"},"schema_version":"1.0","source":{"id":"2604.17376","kind":"arxiv","version":2}},"canonical_sha256":"5912311452e834c613c487dc78b03cf37b7de9631b6f590b74a218d2fbac039b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"5912311452e834c613c487dc78b03cf37b7de9631b6f590b74a218d2fbac039b","first_computed_at":"2026-07-07T03:18:55.918443Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-07T03:18:55.918443Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"hfAgjNYRis5GTIjhXp4HgBKv6gxs318e6+wZ3ZaMTXXpuTtrbAMN+T3Sr/X8j2lbPN0ZHLsGbLo7Tjcj1S9EAA==","signature_status":"signed_v1","signed_at":"2026-07-07T03:18:55.918955Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.17376","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7c84fb99c35c6c4b53878529bc17d6b2c4b743ad70ec96b6767139cfa5e5848b","sha256:fd19d3829b652eefc6a702e244e19ea0de80fa7ff784a82609910ddbd83c7417"],"state_sha256":"286d8540ec5d9f345f2348bc91194d4f02bfdbf53895bc7b9e7e6c1a86cef243"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tMyP+1FAVPwESR79B2RX3EXPJpgqcAec+yhauswz/Jz8rYEmpIVHZD5AaCsf9PXdiF7LiTJ4u7RkHlUZUftpAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T22:09:23.774823Z","bundle_sha256":"8f6561a57e249b505d0cbc6602e79a0d3ee58ed6eee488c019ce867cdda04ee3"}}