{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:RYK24JQPPEHDWU3T2WNWXXR6SE","short_pith_number":"pith:RYK24JQP","canonical_record":{"source":{"id":"2604.27375","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-30T03:39:32Z","cross_cats_sorted":[],"title_canon_sha256":"1351a39195934f95660a3e1b03ef58c391c15fcdbca2b6f19b4e841f675186ec","abstract_canon_sha256":"48d3fba522524a7d5308db1ccacddc9f40322e8a951e3d20a9e35a9e5dd23365"},"schema_version":"1.0"},"canonical_sha256":"8e15ae260f790e3b5373d59b6bde3e9130b41c92eb29a8a2acf5dbdc0c9949cb","source":{"kind":"arxiv","id":"2604.27375","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.27375","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"arxiv_version","alias_value":"2604.27375v2","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.27375","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_12","alias_value":"RYK24JQPPEHD","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_16","alias_value":"RYK24JQPPEHDWU3T","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_8","alias_value":"RYK24JQP","created_at":"2026-05-21T01:05:19Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:RYK24JQPPEHDWU3T2WNWXXR6SE","target":"record","payload":{"canonical_record":{"source":{"id":"2604.27375","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-30T03:39:32Z","cross_cats_sorted":[],"title_canon_sha256":"1351a39195934f95660a3e1b03ef58c391c15fcdbca2b6f19b4e841f675186ec","abstract_canon_sha256":"48d3fba522524a7d5308db1ccacddc9f40322e8a951e3d20a9e35a9e5dd23365"},"schema_version":"1.0"},"canonical_sha256":"8e15ae260f790e3b5373d59b6bde3e9130b41c92eb29a8a2acf5dbdc0c9949cb","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:05:19.666891Z","signature_b64":"JtHGzcOOx4U9MpKVenpeqK7gIrFeGM7DCVGWET5re7uevQnZNvDSNA0dbQWoaT3yNNg8wWH+3YGQzKP8aWLMCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8e15ae260f790e3b5373d59b6bde3e9130b41c92eb29a8a2acf5dbdc0c9949cb","last_reissued_at":"2026-05-21T01:05:19.666166Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:05:19.666166Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.27375","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":"2026-05-21T01:05:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jmqQs0sFyx/MqapmIRTvIlAIxfHpdXeMeqRZGjw1DBa9GVTJXV8ha2yAm/m2sz7f/Iqmq31dcHumdBdviRf4Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T01:46:44.174473Z"},"content_sha256":"6122d44a095fee363e265aea89039c88623c9835271e6eb4ef71b85df0278834","schema_version":"1.0","event_id":"sha256:6122d44a095fee363e265aea89039c88623c9835271e6eb4ef71b85df0278834"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:RYK24JQPPEHDWU3T2WNWXXR6SE","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"VeraRetouch: A Lightweight Fully Differentiable Framework for Multi-Task Reasoning Photo Retouching","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Changqing Zou, Hongliang Wang, Jiajun Tang, Jinwei Chen, Qingnan Fan, Yihong Guo, Yizhuo Zhou, Youwei Lyu","submitted_at":"2026-04-30T03:39:32Z","abstract_excerpt":"Reasoning photo retouching has gained significant traction, requiring models to analyze image defects, give reasoning processes, and execute precise retouching enhancements. However, existing approaches often rely on non-differentiable external software, creating optimization barriers and suffering from high parameter redundancy and limited generalization. To address these challenges, we propose VeraRetouch, a lightweight and fully differentiable framework for multi-task photo retouching. We employ a 0.5B Vision-Language Model (VLM) as the central intelligence to formulate retouching plans bas"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"VeraRetouch achieves state-of-the-art performance across multiple benchmarks while maintaining a significantly smaller footprint, enabling mobile deployment.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The fully differentiable Retouch Renderer can faithfully replicate the effects of professional retouching operations using only decoupled control latents for lighting, global color, and specific colors, without introducing artifacts or losing fidelity compared to external tools.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VeraRetouch is a lightweight fully differentiable framework using a 0.5B VLM for retouching plans and a custom renderer for end-to-end training, backed by a new million-scale dataset and RL post-training, to achieve SOTA multi-task photo retouching with mobile deployment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ffc74d690d8780951a5f016aedc2a262db84ee2d8da83d0f09f3fcd7ef9eb506"},"source":{"id":"2604.27375","kind":"arxiv","version":2},"verdict":{"id":"b740223d-05ae-4333-b464-02321efc589d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T08:55:24.817334Z","strongest_claim":"VeraRetouch achieves state-of-the-art performance across multiple benchmarks while maintaining a significantly smaller footprint, enabling mobile deployment.","one_line_summary":"VeraRetouch is a lightweight fully differentiable framework using a 0.5B VLM for retouching plans and a custom renderer for end-to-end training, backed by a new million-scale dataset and RL post-training, to achieve SOTA multi-task photo retouching with mobile deployment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The fully differentiable Retouch Renderer can faithfully replicate the effects of professional retouching operations using only decoupled control latents for lighting, global color, and specific colors, without introducing artifacts or losing fidelity compared to external tools.","pith_extraction_headline":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.27375/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T22:39:02.626051Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T19:16:37.789145Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"29ef9358c70a28c9b764fd9a2cfedd8097349bbf2e389db8a362de8a2b724d19"},"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":"b740223d-05ae-4333-b464-02321efc589d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-21T01:05:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Ier1e/ANhORv2ayZ0U8tKH29QGYttVRDnxf1ryWCcHG/Nsv6EhoEuIMgNCZYirlgKvy/0G1qNRpv7LkfP7rqDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T01:46:44.174989Z"},"content_sha256":"00892e65c1d01f489329be5a3072dc5f25a95a1c63da8c4304c9d86972718492","schema_version":"1.0","event_id":"sha256:00892e65c1d01f489329be5a3072dc5f25a95a1c63da8c4304c9d86972718492"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/bundle.json","state_url":"https://pith.science/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/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":"2026-06-02T01:46:44Z","links":{"resolver":"https://pith.science/pith/RYK24JQPPEHDWU3T2WNWXXR6SE","bundle":"https://pith.science/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/bundle.json","state":"https://pith.science/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RYK24JQPPEHDWU3T2WNWXXR6SE/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:RYK24JQPPEHDWU3T2WNWXXR6SE","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":"48d3fba522524a7d5308db1ccacddc9f40322e8a951e3d20a9e35a9e5dd23365","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-30T03:39:32Z","title_canon_sha256":"1351a39195934f95660a3e1b03ef58c391c15fcdbca2b6f19b4e841f675186ec"},"schema_version":"1.0","source":{"id":"2604.27375","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.27375","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"arxiv_version","alias_value":"2604.27375v2","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.27375","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_12","alias_value":"RYK24JQPPEHD","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_16","alias_value":"RYK24JQPPEHDWU3T","created_at":"2026-05-21T01:05:19Z"},{"alias_kind":"pith_short_8","alias_value":"RYK24JQP","created_at":"2026-05-21T01:05:19Z"}],"graph_snapshots":[{"event_id":"sha256:00892e65c1d01f489329be5a3072dc5f25a95a1c63da8c4304c9d86972718492","target":"graph","created_at":"2026-05-21T01:05:19Z","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":"VeraRetouch achieves state-of-the-art performance across multiple benchmarks while maintaining a significantly smaller footprint, enabling mobile deployment."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The fully differentiable Retouch Renderer can faithfully replicate the effects of professional retouching operations using only decoupled control latents for lighting, global color, and specific colors, without introducing artifacts or losing fidelity compared to external tools."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"VeraRetouch is a lightweight fully differentiable framework using a 0.5B VLM for retouching plans and a custom renderer for end-to-end training, backed by a new million-scale dataset and RL post-training, to achieve SOTA multi-task photo retouching with mobile deployment."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching."}],"snapshot_sha256":"ffc74d690d8780951a5f016aedc2a262db84ee2d8da83d0f09f3fcd7ef9eb506"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T22:39:02.626051Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T19:16:37.789145Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.27375/integrity.json","findings":[],"snapshot_sha256":"29ef9358c70a28c9b764fd9a2cfedd8097349bbf2e389db8a362de8a2b724d19","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reasoning photo retouching has gained significant traction, requiring models to analyze image defects, give reasoning processes, and execute precise retouching enhancements. However, existing approaches often rely on non-differentiable external software, creating optimization barriers and suffering from high parameter redundancy and limited generalization. To address these challenges, we propose VeraRetouch, a lightweight and fully differentiable framework for multi-task photo retouching. We employ a 0.5B Vision-Language Model (VLM) as the central intelligence to formulate retouching plans bas","authors_text":"Changqing Zou, Hongliang Wang, Jiajun Tang, Jinwei Chen, Qingnan Fan, Yihong Guo, Yizhuo Zhou, Youwei Lyu","cross_cats":[],"headline":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-30T03:39:32Z","title":"VeraRetouch: A Lightweight Fully Differentiable Framework for Multi-Task Reasoning Photo Retouching"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.27375","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-07T08:55:24.817334Z","id":"b740223d-05ae-4333-b464-02321efc589d","model_set":{"reader":"grok-4.3"},"one_line_summary":"VeraRetouch is a lightweight fully differentiable framework using a 0.5B VLM for retouching plans and a custom renderer for end-to-end training, backed by a new million-scale dataset and RL post-training, to achieve SOTA multi-task photo retouching with mobile deployment.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"VeraRetouch replaces non-differentiable external tools with a custom renderer to allow end-to-end training of reasoning photo retouching.","strongest_claim":"VeraRetouch achieves state-of-the-art performance across multiple benchmarks while maintaining a significantly smaller footprint, enabling mobile deployment.","weakest_assumption":"The fully differentiable Retouch Renderer can faithfully replicate the effects of professional retouching operations using only decoupled control latents for lighting, global color, and specific colors, without introducing artifacts or losing fidelity compared to external tools."}},"verdict_id":"b740223d-05ae-4333-b464-02321efc589d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:6122d44a095fee363e265aea89039c88623c9835271e6eb4ef71b85df0278834","target":"record","created_at":"2026-05-21T01:05:19Z","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":"48d3fba522524a7d5308db1ccacddc9f40322e8a951e3d20a9e35a9e5dd23365","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-30T03:39:32Z","title_canon_sha256":"1351a39195934f95660a3e1b03ef58c391c15fcdbca2b6f19b4e841f675186ec"},"schema_version":"1.0","source":{"id":"2604.27375","kind":"arxiv","version":2}},"canonical_sha256":"8e15ae260f790e3b5373d59b6bde3e9130b41c92eb29a8a2acf5dbdc0c9949cb","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8e15ae260f790e3b5373d59b6bde3e9130b41c92eb29a8a2acf5dbdc0c9949cb","first_computed_at":"2026-05-21T01:05:19.666166Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-21T01:05:19.666166Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JtHGzcOOx4U9MpKVenpeqK7gIrFeGM7DCVGWET5re7uevQnZNvDSNA0dbQWoaT3yNNg8wWH+3YGQzKP8aWLMCw==","signature_status":"signed_v1","signed_at":"2026-05-21T01:05:19.666891Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.27375","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6122d44a095fee363e265aea89039c88623c9835271e6eb4ef71b85df0278834","sha256:00892e65c1d01f489329be5a3072dc5f25a95a1c63da8c4304c9d86972718492"],"state_sha256":"956b5bd7741386443ecdf4459ba85435454d8b6d582ed8fadb5c3a82fb5717f4"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"n0L8lyIQRCLmL/f1Jd5mH9aCAMr/f2QLv/V+iEOdgCE31wmk6+hjxxphLvA6njv8OErvitu1B0krAdAlAWupDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-02T01:46:44.177361Z","bundle_sha256":"a516f8b2c559363d58d984a1835a62b22aa72a2f44aef6304b8d0e6ac8dff18d"}}