{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:FFZ7H5JWCY5UQO4EDINX643UWQ","short_pith_number":"pith:FFZ7H5JW","canonical_record":{"source":{"id":"2506.22049","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T09:45:15Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"3d7303e6d6d5cd7e576ed02bbddef849d6b96f2fa5b25f98f684a4f79ed710f9","abstract_canon_sha256":"f16f4d6c1a6d08a4417a31e1880004595badbfb6932079a882785505b2801d61"},"schema_version":"1.0"},"canonical_sha256":"2973f3f536163b483b841a1b7f7374b41bfabe1c6f417be7077bac5a91532762","source":{"kind":"arxiv","id":"2506.22049","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.22049","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"arxiv_version","alias_value":"2506.22049v2","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.22049","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_12","alias_value":"FFZ7H5JWCY5U","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_16","alias_value":"FFZ7H5JWCY5UQO4E","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_8","alias_value":"FFZ7H5JW","created_at":"2026-07-05T11:31:28Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:FFZ7H5JWCY5UQO4EDINX643UWQ","target":"record","payload":{"canonical_record":{"source":{"id":"2506.22049","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T09:45:15Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"3d7303e6d6d5cd7e576ed02bbddef849d6b96f2fa5b25f98f684a4f79ed710f9","abstract_canon_sha256":"f16f4d6c1a6d08a4417a31e1880004595badbfb6932079a882785505b2801d61"},"schema_version":"1.0"},"canonical_sha256":"2973f3f536163b483b841a1b7f7374b41bfabe1c6f417be7077bac5a91532762","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:31:28.941733Z","signature_b64":"ezYO289LSyUv9eQiMP4ufKOT0ge7Slo0eTM6qWHPB6F09T1fpTw5M2SQczvGtTlJQttSy/vKdss1FxmZcjSoBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2973f3f536163b483b841a1b7f7374b41bfabe1c6f417be7077bac5a91532762","last_reissued_at":"2026-07-05T11:31:28.941239Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:31:28.941239Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2506.22049","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":"2026-07-05T11:31:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1oDDO88v9LqRQ37pnbImL/v0o5+Iyu23BEVNcIvNUPxh3+a0jmnBAM8L3li8UG4CkdR/F4gecLmfS9A1kPe6Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T06:02:56.414452Z"},"content_sha256":"e779a3fc4a8b9934453a4a0093f873f12a0222e7644e8326c896641c6bea574d","schema_version":"1.0","event_id":"sha256:e779a3fc4a8b9934453a4a0093f873f12a0222e7644e8326c896641c6bea574d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:FFZ7H5JWCY5UQO4EDINX643UWQ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ajay Kumar Jaiswal, Can Yang, Fan Zhang, Hao Chen, Jishan Hu, Lu Yin, Pengxiang Li, Shiwei Liu, Shizhe Diao, Tianhao Chen, Xin Xu, Xinyuan Song, Yang Wang, Yu Li, Zijing Liu","submitted_at":"2025-06-27T09:45:15Z","abstract_excerpt":"Modern Large Language Models, such as the LLaMA, Qwen and DeepSeek series, predominantly adopt the Pre-LayerNorm (Pre-LN) Transformer architecture. While being stable during pretraining and scalable to large model sizes, Pre-LN suffers from an exponential growth in activation variance across layers, causing the shortcut to dominate over sub-layer outputs in the residual connection and limiting the learning capacity of deeper layers. To mitigate this issue, we propose Gradient-Preserving Activation Scaling (GPAS), a simple technique that can be used in combination with existing approaches. GPAS"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.22049","kind":"arxiv","version":2},"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/2506.22049/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":"2026-07-05T11:31:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"yiivAzeJFX+JaR+nr/0VZxCKhO4IIYFXkM/pzipC4zX/txBrgm6zEFQT93z7nT5KBNz3tWOJlPn/YsMp9NI6BQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T06:02:56.414973Z"},"content_sha256":"7d19c9290585471b7272d07d42f0af2b0f83bd07e043e22c328270d35228e385","schema_version":"1.0","event_id":"sha256:7d19c9290585471b7272d07d42f0af2b0f83bd07e043e22c328270d35228e385"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/bundle.json","state_url":"https://pith.science/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/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":"2026-08-17T06:02:56Z","links":{"resolver":"https://pith.science/pith/FFZ7H5JWCY5UQO4EDINX643UWQ","bundle":"https://pith.science/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/bundle.json","state":"https://pith.science/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FFZ7H5JWCY5UQO4EDINX643UWQ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:FFZ7H5JWCY5UQO4EDINX643UWQ","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":"f16f4d6c1a6d08a4417a31e1880004595badbfb6932079a882785505b2801d61","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T09:45:15Z","title_canon_sha256":"3d7303e6d6d5cd7e576ed02bbddef849d6b96f2fa5b25f98f684a4f79ed710f9"},"schema_version":"1.0","source":{"id":"2506.22049","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.22049","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"arxiv_version","alias_value":"2506.22049v2","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.22049","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_12","alias_value":"FFZ7H5JWCY5U","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_16","alias_value":"FFZ7H5JWCY5UQO4E","created_at":"2026-07-05T11:31:28Z"},{"alias_kind":"pith_short_8","alias_value":"FFZ7H5JW","created_at":"2026-07-05T11:31:28Z"}],"graph_snapshots":[{"event_id":"sha256:7d19c9290585471b7272d07d42f0af2b0f83bd07e043e22c328270d35228e385","target":"graph","created_at":"2026-07-05T11:31:28Z","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/2506.22049/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Modern Large Language Models, such as the LLaMA, Qwen and DeepSeek series, predominantly adopt the Pre-LayerNorm (Pre-LN) Transformer architecture. While being stable during pretraining and scalable to large model sizes, Pre-LN suffers from an exponential growth in activation variance across layers, causing the shortcut to dominate over sub-layer outputs in the residual connection and limiting the learning capacity of deeper layers. To mitigate this issue, we propose Gradient-Preserving Activation Scaling (GPAS), a simple technique that can be used in combination with existing approaches. GPAS","authors_text":"Ajay Kumar Jaiswal, Can Yang, Fan Zhang, Hao Chen, Jishan Hu, Lu Yin, Pengxiang Li, Shiwei Liu, Shizhe Diao, Tianhao Chen, Xin Xu, Xinyuan Song, Yang Wang, Yu Li, Zijing Liu","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T09:45:15Z","title":"GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.22049","kind":"arxiv","version":2},"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:e779a3fc4a8b9934453a4a0093f873f12a0222e7644e8326c896641c6bea574d","target":"record","created_at":"2026-07-05T11:31:28Z","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":"f16f4d6c1a6d08a4417a31e1880004595badbfb6932079a882785505b2801d61","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T09:45:15Z","title_canon_sha256":"3d7303e6d6d5cd7e576ed02bbddef849d6b96f2fa5b25f98f684a4f79ed710f9"},"schema_version":"1.0","source":{"id":"2506.22049","kind":"arxiv","version":2}},"canonical_sha256":"2973f3f536163b483b841a1b7f7374b41bfabe1c6f417be7077bac5a91532762","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2973f3f536163b483b841a1b7f7374b41bfabe1c6f417be7077bac5a91532762","first_computed_at":"2026-07-05T11:31:28.941239Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:31:28.941239Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ezYO289LSyUv9eQiMP4ufKOT0ge7Slo0eTM6qWHPB6F09T1fpTw5M2SQczvGtTlJQttSy/vKdss1FxmZcjSoBA==","signature_status":"signed_v1","signed_at":"2026-07-05T11:31:28.941733Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.22049","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e779a3fc4a8b9934453a4a0093f873f12a0222e7644e8326c896641c6bea574d","sha256:7d19c9290585471b7272d07d42f0af2b0f83bd07e043e22c328270d35228e385"],"state_sha256":"80b0c83bfb547f9d17e1f013d26501a7f74e2cba250e8b2504976f2b5f7ef0c9"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"aw4bg4BQNxPMtw5qJM8TqVO1vwbStIk8j03SHoKY6TvMuCFjxnu7mhCKKZGnJFEzNxcCQb3Ytu+OHbpThFePDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-17T06:02:56.418794Z","bundle_sha256":"91ecded6ec9071908aa6cb5db90a3679598390802d3795722cc009d1be785cbe"}}