{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:UPHQIYP2QOJJVFM2EUGWH7M7RU","short_pith_number":"pith:UPHQIYP2","canonical_record":{"source":{"id":"2302.01172","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-02T15:49:03Z","cross_cats_sorted":[],"title_canon_sha256":"229997eee134f0b9f5dca93b828e94f15008e3193b3682472df6857fc7fee3ea","abstract_canon_sha256":"38296cf27cd298e28f265c29e7825123daf2a08c2c57b091b6ab78546c98aacc"},"schema_version":"1.0"},"canonical_sha256":"a3cf0461fa83929a959a250d63fd9f8d095cbb771b3a34b89bbe87e9ebdbb5c9","source":{"kind":"arxiv","id":"2302.01172","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2302.01172","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"arxiv_version","alias_value":"2302.01172v1","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.01172","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_12","alias_value":"UPHQIYP2QOJJ","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_16","alias_value":"UPHQIYP2QOJJVFM2","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_8","alias_value":"UPHQIYP2","created_at":"2026-07-05T05:38:23Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:UPHQIYP2QOJJVFM2EUGWH7M7RU","target":"record","payload":{"canonical_record":{"source":{"id":"2302.01172","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-02T15:49:03Z","cross_cats_sorted":[],"title_canon_sha256":"229997eee134f0b9f5dca93b828e94f15008e3193b3682472df6857fc7fee3ea","abstract_canon_sha256":"38296cf27cd298e28f265c29e7825123daf2a08c2c57b091b6ab78546c98aacc"},"schema_version":"1.0"},"canonical_sha256":"a3cf0461fa83929a959a250d63fd9f8d095cbb771b3a34b89bbe87e9ebdbb5c9","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:38:23.771628Z","signature_b64":"N3woKFpEzOf7VwMhCix9KfVqwBNZ0rfPKpwF9v1v3FOzqRCv4ujeAC3vYdGXlALkj9sZa1k8iX2GfAF5XJekCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3cf0461fa83929a959a250d63fd9f8d095cbb771b3a34b89bbe87e9ebdbb5c9","last_reissued_at":"2026-07-05T05:38:23.771256Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:38:23.771256Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2302.01172","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:38:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8K4NHhUrQgR/MGDX3ocIS2LGRwNwjajW9mXhlT/rY7GSHcrSmvxwPs7h0HZbFZQPeqCpdr9Aa5/BAddXWNh1BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T09:17:50.662017Z"},"content_sha256":"ffd44ed792664d0ba563f6d2474d6373e8fd633de697775aa6c3665c2b7d51cc","schema_version":"1.0","event_id":"sha256:ffd44ed792664d0ba563f6d2474d6373e8fd633de697775aa6c3665c2b7d51cc"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:UPHQIYP2QOJJVFM2EUGWH7M7RU","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amir Yazdanbakhsh, Christopher De Sa, Oleg Rybakov, Shivani Agrawal, Suvinay Subramanian, Yucheng Lu","submitted_at":"2023-02-02T15:49:03Z","abstract_excerpt":"Recent innovations on hardware (e.g. Nvidia A100) have motivated learning N:M structured sparsity masks from scratch for fast model inference. However, state-of-the-art learning recipes in this regime (e.g. SR-STE) are proposed for non-adaptive optimizers like momentum SGD, while incurring non-trivial accuracy drop for Adam-trained models like attention-based LLMs. In this paper, we first demonstrate such gap origins from poorly estimated second moment (i.e. variance) in Adam states given by the masked weights. We conjecture that learning N:M masks with Adam should take the critical regime of "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.01172","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2302.01172/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":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:38:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GfcsiLLYpFZpcgFZopUs18w/rIC/XKtX5UFjyTfq4tmaYlob8SZ3jiQ4uhTRuSTTtcxjRz/Ulm9fC3Tpus17Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T09:17:50.662541Z"},"content_sha256":"dba0cbe5f1530b1d87032087ad1f7212d925166d9b1195979ff8a8887e70ed98","schema_version":"1.0","event_id":"sha256:dba0cbe5f1530b1d87032087ad1f7212d925166d9b1195979ff8a8887e70ed98"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/bundle.json","state_url":"https://pith.science/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/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":"2026-08-09T09:17:50Z","links":{"resolver":"https://pith.science/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU","bundle":"https://pith.science/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/bundle.json","state":"https://pith.science/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UPHQIYP2QOJJVFM2EUGWH7M7RU/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:UPHQIYP2QOJJVFM2EUGWH7M7RU","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":"38296cf27cd298e28f265c29e7825123daf2a08c2c57b091b6ab78546c98aacc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-02T15:49:03Z","title_canon_sha256":"229997eee134f0b9f5dca93b828e94f15008e3193b3682472df6857fc7fee3ea"},"schema_version":"1.0","source":{"id":"2302.01172","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2302.01172","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"arxiv_version","alias_value":"2302.01172v1","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.01172","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_12","alias_value":"UPHQIYP2QOJJ","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_16","alias_value":"UPHQIYP2QOJJVFM2","created_at":"2026-07-05T05:38:23Z"},{"alias_kind":"pith_short_8","alias_value":"UPHQIYP2","created_at":"2026-07-05T05:38:23Z"}],"graph_snapshots":[{"event_id":"sha256:dba0cbe5f1530b1d87032087ad1f7212d925166d9b1195979ff8a8887e70ed98","target":"graph","created_at":"2026-07-05T05:38:23Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2302.01172/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent innovations on hardware (e.g. Nvidia A100) have motivated learning N:M structured sparsity masks from scratch for fast model inference. However, state-of-the-art learning recipes in this regime (e.g. SR-STE) are proposed for non-adaptive optimizers like momentum SGD, while incurring non-trivial accuracy drop for Adam-trained models like attention-based LLMs. In this paper, we first demonstrate such gap origins from poorly estimated second moment (i.e. variance) in Adam states given by the masked weights. We conjecture that learning N:M masks with Adam should take the critical regime of ","authors_text":"Amir Yazdanbakhsh, Christopher De Sa, Oleg Rybakov, Shivani Agrawal, Suvinay Subramanian, Yucheng Lu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-02T15:49:03Z","title":"STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.01172","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ffd44ed792664d0ba563f6d2474d6373e8fd633de697775aa6c3665c2b7d51cc","target":"record","created_at":"2026-07-05T05:38:23Z","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":"38296cf27cd298e28f265c29e7825123daf2a08c2c57b091b6ab78546c98aacc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-02T15:49:03Z","title_canon_sha256":"229997eee134f0b9f5dca93b828e94f15008e3193b3682472df6857fc7fee3ea"},"schema_version":"1.0","source":{"id":"2302.01172","kind":"arxiv","version":1}},"canonical_sha256":"a3cf0461fa83929a959a250d63fd9f8d095cbb771b3a34b89bbe87e9ebdbb5c9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a3cf0461fa83929a959a250d63fd9f8d095cbb771b3a34b89bbe87e9ebdbb5c9","first_computed_at":"2026-07-05T05:38:23.771256Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T05:38:23.771256Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"N3woKFpEzOf7VwMhCix9KfVqwBNZ0rfPKpwF9v1v3FOzqRCv4ujeAC3vYdGXlALkj9sZa1k8iX2GfAF5XJekCA==","signature_status":"signed_v1","signed_at":"2026-07-05T05:38:23.771628Z","signed_message":"canonical_sha256_bytes"},"source_id":"2302.01172","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ffd44ed792664d0ba563f6d2474d6373e8fd633de697775aa6c3665c2b7d51cc","sha256:dba0cbe5f1530b1d87032087ad1f7212d925166d9b1195979ff8a8887e70ed98"],"state_sha256":"ba0b8309a7515df56b04d10ef06ccb669e587f5ca537db12445aa5fdf0cbf270"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+PdoSmOX9PK17zuf3gJGMhyzz97/omnZ5sXof+9wOlJl4LTl0mZLmjnP1AnSqA1eVrzQulUUNaCIihWnsp9+AA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T09:17:50.666489Z","bundle_sha256":"c84dfc36983500e33ab1976d41c017ca2e4d263fdface9e3dad8d61524c19d15"}}