{"paper":{"title":"Generative Refinement Networks for Visual Synthesis","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Generative Refinement Networks combine near-lossless quantization with global refinement to surpass diffusion and autoregressive models in visual synthesis.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bingyue Peng, Jiahuan Wang, Jian Han, Jinlai Liu, Zehuan Yuan","submitted_at":"2026-04-14T17:59:03Z","abstract_excerpt":"While diffusion models dominate the field of visual generation, they are computationally inefficient, applying a uniform computational effort regardless of different complexity. In contrast, autoregressive (AR) models are inherently complexity-aware, as evidenced by their variable likelihoods, but are often hindered by lossy discrete tokenization and error accumulation. In this work, we introduce Generative Refinement Networks (GRN), a next-generation visual synthesis paradigm that addresses these issues. At its core, GRN addresses the discrete tokenization bottleneck through a theoretically n"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the ImageNet benchmark, GRN establishes new records in image reconstruction (0.56 rFID) and class-conditional image generation (1.81 gFID). We also scale GRN to more challenging text-to-image and text-to-video generation, delivering superior performance on an equivalent scale.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the Hierarchical Binary Quantization is theoretically near-lossless and that the global refinement mechanism corrects errors without introducing new accumulation problems or requiring post-hoc tuning that affects the reported metrics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"GRN uses hierarchical binary quantization and entropy-guided refinement to set new ImageNet records of 0.56 rFID for reconstruction and 1.81 gFID for class-conditional generation while releasing code and models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Generative Refinement Networks combine near-lossless quantization with global refinement to surpass diffusion and autoregressive models in visual synthesis.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7237182ebe3a0838a0eb5990bbc9ea321f13487929f1ffe114e96713404a2a84"},"source":{"id":"2604.13030","kind":"arxiv","version":2},"verdict":{"id":"b00fa9a5-57b3-449e-a0aa-8aa439dde081","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:37:19.049077Z","strongest_claim":"On the ImageNet benchmark, GRN establishes new records in image reconstruction (0.56 rFID) and class-conditional image generation (1.81 gFID). We also scale GRN to more challenging text-to-image and text-to-video generation, delivering superior performance on an equivalent scale.","one_line_summary":"GRN uses hierarchical binary quantization and entropy-guided refinement to set new ImageNet records of 0.56 rFID for reconstruction and 1.81 gFID for class-conditional generation while releasing code and models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the Hierarchical Binary Quantization is theoretically near-lossless and that the global refinement mechanism corrects errors without introducing new accumulation problems or requiring post-hoc tuning that affects the reported metrics.","pith_extraction_headline":"Generative Refinement Networks combine near-lossless quantization with global refinement to surpass diffusion and autoregressive models in visual synthesis."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.13030/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"}