{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:V6FO5VMUUFJSHSFRGJMOFIG3ZP","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":"2d08372d9a4c7b48dbc14e28644d640bf9ae342b1f3921f992fd13232fc40b07","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:52:38Z","title_canon_sha256":"1b015046d7a5cfebfacbfb405d8f00f88db4ab80a6a8aef77987b0e9b4c4f95e"},"schema_version":"1.0","source":{"id":"2605.10989","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.10989","created_at":"2026-05-20T00:00:42Z"},{"alias_kind":"arxiv_version","alias_value":"2605.10989v2","created_at":"2026-05-20T00:00:42Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.10989","created_at":"2026-05-20T00:00:42Z"},{"alias_kind":"pith_short_12","alias_value":"V6FO5VMUUFJS","created_at":"2026-05-20T00:00:42Z"},{"alias_kind":"pith_short_16","alias_value":"V6FO5VMUUFJSHSFR","created_at":"2026-05-20T00:00:42Z"},{"alias_kind":"pith_short_8","alias_value":"V6FO5VMU","created_at":"2026-05-20T00:00:42Z"}],"graph_snapshots":[{"event_id":"sha256:03b6e2dc503a3a9d770cd93986089b1ac44f9138ce3c4d03aef550e51c37494c","target":"graph","created_at":"2026-05-20T00:00:42Z","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":"Experiments on image classification, object detection, and language understanding tasks demonstrate that SURGE performs best over state-of-the-art methods."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The full-precision auxiliary branch in DPGC provides bias-reduced gradient estimates beyond the first-order STE approximation without introducing new mismatches or instabilities during training."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SURGE introduces a dual-path gradient compensator and adaptive scaler to improve surrogate gradient estimation in binarized neural network training."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Binary neural networks train more accurately when a learnable surrogate gradient uses an auxiliary full-precision branch to reduce mismatch."}],"snapshot_sha256":"72687c3fdfd7539a13740308d6a9c557b2a849fc4d6e9e71a8704cd4f1d147b9"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T22:33:54.938202Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T14:01:21.770622Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:42:35.079079Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.10989/integrity.json","findings":[],"snapshot_sha256":"ca90ec4a1b3b3cc7c3c3f17fcbe07abbab950d61e669f743cb4e3207234eef66","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The training of Binary Neural Networks (BNNs) is fundamentally based on gradient approximation for non-differentiable binarization operations (e.g., sign function). However, prevailing methods including the Straight-Through Estimator (STE) and its improved variants, rely on hand-crafted designs that suffer from gradient mismatch problem and information loss induced by fixed-range gradient clipping. To address this, we propose SURrogate GradiEnt Adaptation (SURGE), a novel learnable gradient compensation framework with theoretical grounding. SURGE mitigates gradient mismatch through auxiliary b","authors_text":"Baochang Zhang, Boyu Liu, Canyu Chen, Haoyu Huang, Linlin Yang, Xuhui Liu, Yanjing Li, Yuguang Yang, Zhongqian Fu","cross_cats":["cs.AI"],"headline":"Binary neural networks train more accurately when a learnable surrogate gradient uses an auxiliary full-precision branch to reduce mismatch.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:52:38Z","title":"SURGE: Surrogate Gradient Adaptation in Binary Neural Networks"},"references":{"count":129,"internal_anchors":11,"resolved_work":129,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Scaling Learning Algorithms Towards","work_id":"bb2761cc-98d0-411b-92f6-803773d64460","year":null},{"cited_arxiv_id":"1308.3432","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation","work_id":"1fe8c7c8-aff7-4b94-9096-e549d7e60789","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Learning sparse neural networks through L\\_0 regularization , author=","work_id":"3379164a-b285-46fb-a1f7-e2faa40cab45","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Differentiable soft quantization: Bridging full-precision and low-bit neural networks , author=","work_id":"f1feb016-058a-4be2-8e69-a25098204a0b","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Binaryconnect: Training deep neural networks with binary weights during propagations , author=","work_id":"e0aa0686-1e81-4f1a-88bd-3e01e1d53ed5","year":null}],"snapshot_sha256":"5db2ec13f37ae26fc2e739c2b242a921a7b04b642c4041d8465b906edfc93dfb"},"source":{"id":"2605.10989","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-19T17:45:15.360737Z","id":"bc5bbb6d-3359-4593-8909-ecb642af323e","model_set":{"reader":"grok-4.3"},"one_line_summary":"SURGE introduces a dual-path gradient compensator and adaptive scaler to improve surrogate gradient estimation in binarized neural network training.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Binary neural networks train more accurately when a learnable surrogate gradient uses an auxiliary full-precision branch to reduce mismatch.","strongest_claim":"Experiments on image classification, object detection, and language understanding tasks demonstrate that SURGE performs best over state-of-the-art methods.","weakest_assumption":"The full-precision auxiliary branch in DPGC provides bias-reduced gradient estimates beyond the first-order STE approximation without introducing new mismatches or instabilities during training."}},"verdict_id":"bc5bbb6d-3359-4593-8909-ecb642af323e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c0ce83359a5aaeceb91184476af6d6edc4f7aef35f56215a6d185f1c144e0a6f","target":"record","created_at":"2026-05-20T00:00:42Z","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":"2d08372d9a4c7b48dbc14e28644d640bf9ae342b1f3921f992fd13232fc40b07","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:52:38Z","title_canon_sha256":"1b015046d7a5cfebfacbfb405d8f00f88db4ab80a6a8aef77987b0e9b4c4f95e"},"schema_version":"1.0","source":{"id":"2605.10989","kind":"arxiv","version":2}},"canonical_sha256":"af8aeed594a15323c8b13258e2a0dbcbf282fef13b828e8611a498acd564b864","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"af8aeed594a15323c8b13258e2a0dbcbf282fef13b828e8611a498acd564b864","first_computed_at":"2026-05-20T00:00:42.650092Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:42.650092Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Jqul8IKxidIPJXRUoHA3Hw2maaxYpI3ZDfLDsXz5dszJRW/JtGRA/c9dHLYmcaMox2KZhcwX2gbNwAvtUqUHDQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:42.650843Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.10989","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c0ce83359a5aaeceb91184476af6d6edc4f7aef35f56215a6d185f1c144e0a6f","sha256:03b6e2dc503a3a9d770cd93986089b1ac44f9138ce3c4d03aef550e51c37494c"],"state_sha256":"d5713b037978755c95dbed380ff9e08dbee270626ca2b6a1174160174ee4b47c"}