{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OTAJLRKWGWPABL6SIZK3MQKK3V","short_pith_number":"pith:OTAJLRKW","schema_version":"1.0","canonical_sha256":"74c095c556359e00afd24655b6414add74a8e59fa8691ee43cd2ebc82e9ffa52","source":{"kind":"arxiv","id":"2505.19342","version":2},"attestation_state":"computed","paper":{"title":"ASTRA: Communication-Efficient Acceleration for Multi-Device Transformer Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Deepak Ganesan, Hui Guan, Lijun Zhang, Xiao Liu","submitted_at":"2025-05-25T22:16:59Z","abstract_excerpt":"Multi-device inference can reduce Transformer latency by parallelizing computation. However, existing methods require high inter-device bandwidth, making them impractical for bandwidth-constrained environments. We present ASTRA, a communication-efficient framework that integrates sequence parallelism with mixed-precision attention, where non-local token embeddings are transmitted as low-bit vector-quantized codes while local attention remains full precision. To preserve accuracy under aggressive compression, ASTRA introduces Noise-Augmented Quantization and Distributed Class Tokens. Across vis"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2505.19342","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T22:16:59Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c854ac4c837e8b001a185a257248b9dd7fcc3a4117aed362d098c52304de2996","abstract_canon_sha256":"665ffff21f65346ad1fa276b5bf59d88febc7a2fe278dcfd60d08e7dc6c68650"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:28.301965Z","signature_b64":"Zb8+bcheBF3J29Z10UDZUSiOMfxMJ5/Gu3eAQoArE0K/v4kgfezRVNOzRLd68nEdxaLSoKqGkLEuOLCQB2N6BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"74c095c556359e00afd24655b6414add74a8e59fa8691ee43cd2ebc82e9ffa52","last_reissued_at":"2026-05-28T01:04:28.301381Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:28.301381Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ASTRA: Communication-Efficient Acceleration for Multi-Device Transformer Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Deepak Ganesan, Hui Guan, Lijun Zhang, Xiao Liu","submitted_at":"2025-05-25T22:16:59Z","abstract_excerpt":"Multi-device inference can reduce Transformer latency by parallelizing computation. However, existing methods require high inter-device bandwidth, making them impractical for bandwidth-constrained environments. We present ASTRA, a communication-efficient framework that integrates sequence parallelism with mixed-precision attention, where non-local token embeddings are transmitted as low-bit vector-quantized codes while local attention remains full precision. To preserve accuracy under aggressive compression, ASTRA introduces Noise-Augmented Quantization and Distributed Class Tokens. Across vis"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.19342","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/2505.19342/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2505.19342","created_at":"2026-05-28T01:04:28.301450+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.19342v2","created_at":"2026-05-28T01:04:28.301450+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.19342","created_at":"2026-05-28T01:04:28.301450+00:00"},{"alias_kind":"pith_short_12","alias_value":"OTAJLRKWGWPA","created_at":"2026-05-28T01:04:28.301450+00:00"},{"alias_kind":"pith_short_16","alias_value":"OTAJLRKWGWPABL6S","created_at":"2026-05-28T01:04:28.301450+00:00"},{"alias_kind":"pith_short_8","alias_value":"OTAJLRKW","created_at":"2026-05-28T01:04:28.301450+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.15694","citing_title":"Going Beyond the Edge: Distributed Inference of Transformer Models on Ultra-Low-Power Wireless Devices","ref_index":26,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V","json":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V.json","graph_json":"https://pith.science/api/pith-number/OTAJLRKWGWPABL6SIZK3MQKK3V/graph.json","events_json":"https://pith.science/api/pith-number/OTAJLRKWGWPABL6SIZK3MQKK3V/events.json","paper":"https://pith.science/paper/OTAJLRKW"},"agent_actions":{"view_html":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V","download_json":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V.json","view_paper":"https://pith.science/paper/OTAJLRKW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.19342&json=true","fetch_graph":"https://pith.science/api/pith-number/OTAJLRKWGWPABL6SIZK3MQKK3V/graph.json","fetch_events":"https://pith.science/api/pith-number/OTAJLRKWGWPABL6SIZK3MQKK3V/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V/action/storage_attestation","attest_author":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V/action/author_attestation","sign_citation":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V/action/citation_signature","submit_replication":"https://pith.science/pith/OTAJLRKWGWPABL6SIZK3MQKK3V/action/replication_record"}},"created_at":"2026-05-28T01:04:28.301450+00:00","updated_at":"2026-05-28T01:04:28.301450+00:00"}