{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GGLR7ENLWSHFC2ULIPVAJBYNJR","short_pith_number":"pith:GGLR7ENL","schema_version":"1.0","canonical_sha256":"31971f91abb48e516a8b43ea04870d4c5eb0d199ca83e901b1efac6fa6e3c16a","source":{"kind":"arxiv","id":"2506.23025","version":1},"attestation_state":"computed","paper":{"title":"Spectra 1.1: Scaling Laws and Efficient Inference for Ternary Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Ayush Kaushal, Francis Couture Harpin, Irina Rish, Majid Behbahani, Prashant Shishodia, Tejas Vaidhya, Vineet Jain, Yuriy Nevmyvaka","submitted_at":"2025-06-28T22:13:43Z","abstract_excerpt":"Large language models (LLMs) are increasingly used across research and industry applications, yet their inference efficiency remains a significant challenge. As the computational power of modern GPU architectures continuously improves, their memory bandwidth and capacity have not scaled proportionally, creating a critical bottleneck during inference. To address this, we investigate ternary language models (TriLMs) that employ quantization-aware training to significantly reduce memory requirements. We first analyze the scalability of TriLMs by conducting a scaling law analysis, revealing that T"},"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":"2506.23025","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-28T22:13:43Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9586529369b8e4639df8a30f5e501228d471970fb01325223206df3052bfd9b0","abstract_canon_sha256":"27b6432f96c04432d6dfe9eb3a57cee1026edb0be7c3538b0d04dc96be3bd492"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:28:49.392695Z","signature_b64":"zLv8NcPrbl9IJN4Gr1fRu5/pkvxrlSpooZUfg9lHT2+lKrOZQs/aROa61i6GbSryqjKV54u5XrSr7+t1VT6ACA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31971f91abb48e516a8b43ea04870d4c5eb0d199ca83e901b1efac6fa6e3c16a","last_reissued_at":"2026-07-05T11:28:49.392195Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:28:49.392195Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Spectra 1.1: Scaling Laws and Efficient Inference for Ternary Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Ayush Kaushal, Francis Couture Harpin, Irina Rish, Majid Behbahani, Prashant Shishodia, Tejas Vaidhya, Vineet Jain, Yuriy Nevmyvaka","submitted_at":"2025-06-28T22:13:43Z","abstract_excerpt":"Large language models (LLMs) are increasingly used across research and industry applications, yet their inference efficiency remains a significant challenge. As the computational power of modern GPU architectures continuously improves, their memory bandwidth and capacity have not scaled proportionally, creating a critical bottleneck during inference. To address this, we investigate ternary language models (TriLMs) that employ quantization-aware training to significantly reduce memory requirements. We first analyze the scalability of TriLMs by conducting a scaling law analysis, revealing that T"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.23025","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/2506.23025/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":"2506.23025","created_at":"2026-07-05T11:28:49.392253+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.23025v1","created_at":"2026-07-05T11:28:49.392253+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.23025","created_at":"2026-07-05T11:28:49.392253+00:00"},{"alias_kind":"pith_short_12","alias_value":"GGLR7ENLWSHF","created_at":"2026-07-05T11:28:49.392253+00:00"},{"alias_kind":"pith_short_16","alias_value":"GGLR7ENLWSHFC2UL","created_at":"2026-07-05T11:28:49.392253+00:00"},{"alias_kind":"pith_short_8","alias_value":"GGLR7ENL","created_at":"2026-07-05T11:28:49.392253+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.29705","citing_title":"BitTP: The Lightweight Trajectory Prediction Model with BitLLM for Edge-Devices","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR","json":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR.json","graph_json":"https://pith.science/api/pith-number/GGLR7ENLWSHFC2ULIPVAJBYNJR/graph.json","events_json":"https://pith.science/api/pith-number/GGLR7ENLWSHFC2ULIPVAJBYNJR/events.json","paper":"https://pith.science/paper/GGLR7ENL"},"agent_actions":{"view_html":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR","download_json":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR.json","view_paper":"https://pith.science/paper/GGLR7ENL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.23025&json=true","fetch_graph":"https://pith.science/api/pith-number/GGLR7ENLWSHFC2ULIPVAJBYNJR/graph.json","fetch_events":"https://pith.science/api/pith-number/GGLR7ENLWSHFC2ULIPVAJBYNJR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR/action/storage_attestation","attest_author":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR/action/author_attestation","sign_citation":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR/action/citation_signature","submit_replication":"https://pith.science/pith/GGLR7ENLWSHFC2ULIPVAJBYNJR/action/replication_record"}},"created_at":"2026-07-05T11:28:49.392253+00:00","updated_at":"2026-07-05T11:28:49.392253+00:00"}