{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:24ZDM3RTXEI3ALCEGV4TGEMPRM","short_pith_number":"pith:24ZDM3RT","schema_version":"1.0","canonical_sha256":"d732366e33b911b02c44357933118f8b0ba947778cadbcb85fae0936b3329932","source":{"kind":"arxiv","id":"2407.12327","version":5},"attestation_state":"computed","paper":{"title":"Spectra: Surprising Effectiveness of Pretraining Ternary Language Models at Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Aaryan Bhagat, Arnab Kumar Mondal, Ayush Kaushal, Irina Rish, Tejas Pandey, Tejas Vaidhya","submitted_at":"2024-07-17T05:53:20Z","abstract_excerpt":"Rapid advancements in GPU computational power has outpaced memory capacity and bandwidth growth, creating bottlenecks in Large Language Model (LLM) inference. Post-training quantization is the leading method for addressing memory-related bottlenecks in LLM inference, but it suffers from significant performance degradation below 4-bit precision. This paper addresses these challenges by investigating the pretraining of low-bitwidth models specifically Ternary Language Models (TriLMs) as an alternative to traditional floating-point models (FloatLMs) and their post-training quantized versions (Qua"},"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":"2407.12327","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-17T05:53:20Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"4f3ffb7bda695ecd6f057fc9495a6db82e266c56cee9d2ae1efa2b002af2af5f","abstract_canon_sha256":"cbc97d0dd86eaa601c77328d1faa1a25ae00ad91aa8467a39a0b3f79c939d47d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:18:53.291884Z","signature_b64":"aldMooCwVJuiGUSsYujrM1r81O1sjF2z0u16yQZa3r7tYeysgzbpff/VW1RrPZEIV86N/N6PMo8HNyFBLPJ3Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d732366e33b911b02c44357933118f8b0ba947778cadbcb85fae0936b3329932","last_reissued_at":"2026-07-05T09:18:53.291289Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:18:53.291289Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Spectra: Surprising Effectiveness of Pretraining Ternary Language Models at Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Aaryan Bhagat, Arnab Kumar Mondal, Ayush Kaushal, Irina Rish, Tejas Pandey, Tejas Vaidhya","submitted_at":"2024-07-17T05:53:20Z","abstract_excerpt":"Rapid advancements in GPU computational power has outpaced memory capacity and bandwidth growth, creating bottlenecks in Large Language Model (LLM) inference. Post-training quantization is the leading method for addressing memory-related bottlenecks in LLM inference, but it suffers from significant performance degradation below 4-bit precision. This paper addresses these challenges by investigating the pretraining of low-bitwidth models specifically Ternary Language Models (TriLMs) as an alternative to traditional floating-point models (FloatLMs) and their post-training quantized versions (Qua"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.12327","kind":"arxiv","version":5},"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/2407.12327/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":"2407.12327","created_at":"2026-07-05T09:18:53.291369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.12327v5","created_at":"2026-07-05T09:18:53.291369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.12327","created_at":"2026-07-05T09:18:53.291369+00:00"},{"alias_kind":"pith_short_12","alias_value":"24ZDM3RTXEI3","created_at":"2026-07-05T09:18:53.291369+00:00"},{"alias_kind":"pith_short_16","alias_value":"24ZDM3RTXEI3ALCE","created_at":"2026-07-05T09:18:53.291369+00:00"},{"alias_kind":"pith_short_8","alias_value":"24ZDM3RT","created_at":"2026-07-05T09:18:53.291369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.06485","citing_title":"Litespark Inference For CPUs: Ultra-Fast SIMD Framework for Ternary (1.58-bit) Language Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2512.06443","citing_title":"Vec-LUT: Vector Table Lookup for Parallel Ultra-Low-Bit LLM Inference on Edge Devices","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13859","citing_title":"BiSpikCLM: A Spiking Language Model integrating Softmax-Free Spiking Attention and Spike-Aware Alignment Distillation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25183","citing_title":"Hardware Generation and Exploration of Lookup Table-Based Accelerators for 1.58-bit LLM Inference","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06485","citing_title":"Litespark Inference For CPUs: Ultra-Fast SIMD Framework for Ternary (1.58-bit) Language Models","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM","json":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM.json","graph_json":"https://pith.science/api/pith-number/24ZDM3RTXEI3ALCEGV4TGEMPRM/graph.json","events_json":"https://pith.science/api/pith-number/24ZDM3RTXEI3ALCEGV4TGEMPRM/events.json","paper":"https://pith.science/paper/24ZDM3RT"},"agent_actions":{"view_html":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM","download_json":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM.json","view_paper":"https://pith.science/paper/24ZDM3RT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.12327&json=true","fetch_graph":"https://pith.science/api/pith-number/24ZDM3RTXEI3ALCEGV4TGEMPRM/graph.json","fetch_events":"https://pith.science/api/pith-number/24ZDM3RTXEI3ALCEGV4TGEMPRM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM/action/storage_attestation","attest_author":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM/action/author_attestation","sign_citation":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM/action/citation_signature","submit_replication":"https://pith.science/pith/24ZDM3RTXEI3ALCEGV4TGEMPRM/action/replication_record"}},"created_at":"2026-07-05T09:18:53.291369+00:00","updated_at":"2026-07-05T09:18:53.291369+00:00"}