{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:6OSZOQECNY7TTUZYNCGURXAPOV","short_pith_number":"pith:6OSZOQEC","canonical_record":{"source":{"id":"2602.14068","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-15T09:36:54Z","cross_cats_sorted":[],"title_canon_sha256":"0c004792187db4383cc0de2ca7171e488552a43f3daa5cdff31c367efc1606d6","abstract_canon_sha256":"7b2de2b53e3fd9522674b60b9035b389cf823f5584b88d1c0b2a1c50bec4d98d"},"schema_version":"1.0"},"canonical_sha256":"f3a59740826e3f39d338688d48dc0f754342b8328ceb3a90c7b3b17d3f39b01b","source":{"kind":"arxiv","id":"2602.14068","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.14068","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"arxiv_version","alias_value":"2602.14068v2","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.14068","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"pith_short_12","alias_value":"6OSZOQECNY7T","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6OSZOQECNY7TTUZY","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6OSZOQEC","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:6OSZOQECNY7TTUZYNCGURXAPOV","target":"record","payload":{"canonical_record":{"source":{"id":"2602.14068","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-15T09:36:54Z","cross_cats_sorted":[],"title_canon_sha256":"0c004792187db4383cc0de2ca7171e488552a43f3daa5cdff31c367efc1606d6","abstract_canon_sha256":"7b2de2b53e3fd9522674b60b9035b389cf823f5584b88d1c0b2a1c50bec4d98d"},"schema_version":"1.0"},"canonical_sha256":"f3a59740826e3f39d338688d48dc0f754342b8328ceb3a90c7b3b17d3f39b01b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:16.146186Z","signature_b64":"7O6E77cd2dDXJQadFNGc1Qw7IdIxF1pvxmvNwlsyejiFtYIra4JUl+L9A9IUk+62JNeeHcm8Aw0gDF29PrF0CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f3a59740826e3f39d338688d48dc0f754342b8328ceb3a90c7b3b17d3f39b01b","last_reissued_at":"2026-05-17T23:39:16.145400Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:16.145400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2602.14068","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":"2026-05-17T23:39:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xR+aSgG3wYah1AQcjxSdaSIP6NZC1Q60w9Z/4VnF92QVoR2GAKjR1t34plkHtLe6+5oz5/vgEs2TRONGLNBcDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T08:41:39.059535Z"},"content_sha256":"a14a27ff83bf8357dfdc6062842778dc6619c0037f026ad1b7a270b21ca9e5ae","schema_version":"1.0","event_id":"sha256:a14a27ff83bf8357dfdc6062842778dc6619c0037f026ad1b7a270b21ca9e5ae"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:6OSZOQECNY7TTUZYNCGURXAPOV","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"CoCoEdit: Content-Consistent Image Editing via Region Regularized Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chenxi Xie, Lei Zhang, Liyi Chen, Qiaosi Yi, Ruibin Li, Yuhui Wu","submitted_at":"2026-02-15T09:36:54Z","abstract_excerpt":"Image editing has achieved impressive results with the development of large-scale generative models. However, existing models mainly focus on the editing effects of intended objects and regions, often leading to unwanted changes in unintended regions. We present a post-training framework for Content-Consistent Editing (CoCoEdit) via region regularized reinforcement learning. We first augment existing editing datasets with refined instructions and masks, from which 40K diverse and high quality samples are curated as training set. We then introduce a pixel-level similarity reward to complement M"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Applying CoCoEdit to Qwen-Image-Edit and FLUX-Kontext, we achieve not only competitive editing scores with state-of-the-art models, but also significantly better content consistency, measured by PSNR/SSIM metrics and human subjective ratings.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The region-based regularizer successfully balances preservation of non-edited areas with editing strength without introducing new artifacts or degrading overall quality, relying on the assumption that the combined reward signals accurately reflect desired behavior across diverse images.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CoCoEdit applies region-regularized RL with pixel similarity and MLLM rewards to achieve competitive editing quality alongside significantly improved content consistency on models like Qwen-Image-Edit and FLUX-Kontext.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4988eeb2db904ff7e368994ba1fcb720a08de83d16d03e3245bf6a608ab52259"},"source":{"id":"2602.14068","kind":"arxiv","version":2},"verdict":{"id":"3b70eced-466d-4d30-a1fa-66f84e6a9bf8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T21:46:02.813181Z","strongest_claim":"Applying CoCoEdit to Qwen-Image-Edit and FLUX-Kontext, we achieve not only competitive editing scores with state-of-the-art models, but also significantly better content consistency, measured by PSNR/SSIM metrics and human subjective ratings.","one_line_summary":"CoCoEdit applies region-regularized RL with pixel similarity and MLLM rewards to achieve competitive editing quality alongside significantly improved content consistency on models like Qwen-Image-Edit and FLUX-Kontext.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The region-based regularizer successfully balances preservation of non-edited areas with editing strength without introducing new artifacts or degrading overall quality, relying on the assumption that the combined reward signals accurately reflect desired behavior across diverse images.","pith_extraction_headline":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality."},"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":"3b70eced-466d-4d30-a1fa-66f84e6a9bf8"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6CRLq9tgIxpEJWLJjgJpAaAgFmh1QbxxABSJNATv+nsLgXAgC71XRN9LXYNZ6zMcidh7Sk7s+ZSy3Q8q0e3qBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T08:41:39.059976Z"},"content_sha256":"f2b84cdf70b996769bce5e9ea0da4b1a8d8100ab181992eb5f148dc01e0a846e","schema_version":"1.0","event_id":"sha256:f2b84cdf70b996769bce5e9ea0da4b1a8d8100ab181992eb5f148dc01e0a846e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/6OSZOQECNY7TTUZYNCGURXAPOV/bundle.json","state_url":"https://pith.science/pith/6OSZOQECNY7TTUZYNCGURXAPOV/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/6OSZOQECNY7TTUZYNCGURXAPOV/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":"2026-05-28T08:41:39Z","links":{"resolver":"https://pith.science/pith/6OSZOQECNY7TTUZYNCGURXAPOV","bundle":"https://pith.science/pith/6OSZOQECNY7TTUZYNCGURXAPOV/bundle.json","state":"https://pith.science/pith/6OSZOQECNY7TTUZYNCGURXAPOV/state.json","well_known_bundle":"https://pith.science/.well-known/pith/6OSZOQECNY7TTUZYNCGURXAPOV/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:6OSZOQECNY7TTUZYNCGURXAPOV","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":"7b2de2b53e3fd9522674b60b9035b389cf823f5584b88d1c0b2a1c50bec4d98d","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-15T09:36:54Z","title_canon_sha256":"0c004792187db4383cc0de2ca7171e488552a43f3daa5cdff31c367efc1606d6"},"schema_version":"1.0","source":{"id":"2602.14068","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.14068","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"arxiv_version","alias_value":"2602.14068v2","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.14068","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"pith_short_12","alias_value":"6OSZOQECNY7T","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6OSZOQECNY7TTUZY","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6OSZOQEC","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:f2b84cdf70b996769bce5e9ea0da4b1a8d8100ab181992eb5f148dc01e0a846e","target":"graph","created_at":"2026-05-17T23:39:16Z","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":"Applying CoCoEdit to Qwen-Image-Edit and FLUX-Kontext, we achieve not only competitive editing scores with state-of-the-art models, but also significantly better content consistency, measured by PSNR/SSIM metrics and human subjective ratings."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The region-based regularizer successfully balances preservation of non-edited areas with editing strength without introducing new artifacts or degrading overall quality, relying on the assumption that the combined reward signals accurately reflect desired behavior across diverse images."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"CoCoEdit applies region-regularized RL with pixel similarity and MLLM rewards to achieve competitive editing quality alongside significantly improved content consistency on models like Qwen-Image-Edit and FLUX-Kontext."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality."}],"snapshot_sha256":"4988eeb2db904ff7e368994ba1fcb720a08de83d16d03e3245bf6a608ab52259"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Image editing has achieved impressive results with the development of large-scale generative models. However, existing models mainly focus on the editing effects of intended objects and regions, often leading to unwanted changes in unintended regions. We present a post-training framework for Content-Consistent Editing (CoCoEdit) via region regularized reinforcement learning. We first augment existing editing datasets with refined instructions and masks, from which 40K diverse and high quality samples are curated as training set. We then introduce a pixel-level similarity reward to complement M","authors_text":"Chenxi Xie, Lei Zhang, Liyi Chen, Qiaosi Yi, Ruibin Li, Yuhui Wu","cross_cats":[],"headline":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-15T09:36:54Z","title":"CoCoEdit: Content-Consistent Image Editing via Region Regularized Reinforcement Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.14068","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T21:46:02.813181Z","id":"3b70eced-466d-4d30-a1fa-66f84e6a9bf8","model_set":{"reader":"grok-4.3"},"one_line_summary":"CoCoEdit applies region-regularized RL with pixel similarity and MLLM rewards to achieve competitive editing quality alongside significantly improved content consistency on models like Qwen-Image-Edit and FLUX-Kontext.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Region regularized reinforcement learning trains image editing models to preserve non-edited areas while maintaining edit quality.","strongest_claim":"Applying CoCoEdit to Qwen-Image-Edit and FLUX-Kontext, we achieve not only competitive editing scores with state-of-the-art models, but also significantly better content consistency, measured by PSNR/SSIM metrics and human subjective ratings.","weakest_assumption":"The region-based regularizer successfully balances preservation of non-edited areas with editing strength without introducing new artifacts or degrading overall quality, relying on the assumption that the combined reward signals accurately reflect desired behavior across diverse images."}},"verdict_id":"3b70eced-466d-4d30-a1fa-66f84e6a9bf8"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:a14a27ff83bf8357dfdc6062842778dc6619c0037f026ad1b7a270b21ca9e5ae","target":"record","created_at":"2026-05-17T23:39:16Z","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":"7b2de2b53e3fd9522674b60b9035b389cf823f5584b88d1c0b2a1c50bec4d98d","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-15T09:36:54Z","title_canon_sha256":"0c004792187db4383cc0de2ca7171e488552a43f3daa5cdff31c367efc1606d6"},"schema_version":"1.0","source":{"id":"2602.14068","kind":"arxiv","version":2}},"canonical_sha256":"f3a59740826e3f39d338688d48dc0f754342b8328ceb3a90c7b3b17d3f39b01b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f3a59740826e3f39d338688d48dc0f754342b8328ceb3a90c7b3b17d3f39b01b","first_computed_at":"2026-05-17T23:39:16.145400Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:16.145400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7O6E77cd2dDXJQadFNGc1Qw7IdIxF1pvxmvNwlsyejiFtYIra4JUl+L9A9IUk+62JNeeHcm8Aw0gDF29PrF0CA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:16.146186Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.14068","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a14a27ff83bf8357dfdc6062842778dc6619c0037f026ad1b7a270b21ca9e5ae","sha256:f2b84cdf70b996769bce5e9ea0da4b1a8d8100ab181992eb5f148dc01e0a846e"],"state_sha256":"4dcfebd6946ba35fa7b8117f541601ad93d9ce0635acffedfe9f860be93e85ed"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PH3Iq27eWLrQOAT0dQnuLUZUR/MqWgTxoS8ocD8wWVSVSwIKqgDpK2a4QP4gs2RJ2evMrq38NoUNF+Zse0nWDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-28T08:41:39.062327Z","bundle_sha256":"a2e2b0a9a53ba174260f29d6a181646f58d56ae1e8392938ab3dd4dd6f563aa0"}}