{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XEW3BAY3X5IMYZZ66UOJKNBEXJ","short_pith_number":"pith:XEW3BAY3","schema_version":"1.0","canonical_sha256":"b92db0831bbf50cc673ef51c953424ba425820fa38403c48c94d9db44e976db7","source":{"kind":"arxiv","id":"2505.10475","version":1},"attestation_state":"computed","paper":{"title":"Parallel Scaling Law for Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Binyuan Hui, Dayiheng Liu, Jianling Sun, Jiaxi Yang, Junyang Lin, Mouxiang Chen, Zeyu Cui, Zhongxin Liu","submitted_at":"2025-05-15T16:24:45Z","abstract_excerpt":"It is commonly believed that scaling language models should commit a significant space or time cost, by increasing the parameters (parameter scaling) or output tokens (inference-time scaling). We introduce the third and more inference-efficient scaling paradigm: increasing the model's parallel computation during both training and inference time. We apply $P$ diverse and learnable transformations to the input, execute forward passes of the model in parallel, and dynamically aggregate the $P$ outputs. This method, namely parallel scaling (ParScale), scales parallel computation by reusing existin"},"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":"2505.10475","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-15T16:24:45Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"937079f891024307c03a031e748059f90f694cb11675b0a337e9149d6db89666","abstract_canon_sha256":"3ec47464e6bdca83f843a7085b72088e547704f0ae212f42bf4a4013bb2d19aa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:03:38.818304Z","signature_b64":"b6bnz9iwg6vXQSohLleIPHRKru+7xrexzSF145T+SQyrKP/52waYfRyKFmDoyxaWP+FZuBiPC194nWJKOZYlBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b92db0831bbf50cc673ef51c953424ba425820fa38403c48c94d9db44e976db7","last_reissued_at":"2026-07-05T11:03:38.817825Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:03:38.817825Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Parallel Scaling Law for Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Binyuan Hui, Dayiheng Liu, Jianling Sun, Jiaxi Yang, Junyang Lin, Mouxiang Chen, Zeyu Cui, Zhongxin Liu","submitted_at":"2025-05-15T16:24:45Z","abstract_excerpt":"It is commonly believed that scaling language models should commit a significant space or time cost, by increasing the parameters (parameter scaling) or output tokens (inference-time scaling). We introduce the third and more inference-efficient scaling paradigm: increasing the model's parallel computation during both training and inference time. We apply $P$ diverse and learnable transformations to the input, execute forward passes of the model in parallel, and dynamically aggregate the $P$ outputs. This method, namely parallel scaling (ParScale), scales parallel computation by reusing existin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.10475","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/2505.10475/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":"2505.10475","created_at":"2026-07-05T11:03:38.817883+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.10475v1","created_at":"2026-07-05T11:03:38.817883+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.10475","created_at":"2026-07-05T11:03:38.817883+00:00"},{"alias_kind":"pith_short_12","alias_value":"XEW3BAY3X5IM","created_at":"2026-07-05T11:03:38.817883+00:00"},{"alias_kind":"pith_short_16","alias_value":"XEW3BAY3X5IMYZZ6","created_at":"2026-07-05T11:03:38.817883+00:00"},{"alias_kind":"pith_short_8","alias_value":"XEW3BAY3","created_at":"2026-07-05T11:03:38.817883+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.08696","citing_title":"Structured Recurrent Mixers for Massively Parallelized Sequence Generation","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08696","citing_title":"Structured Recurrent Mixers for Massively Parallelized Sequence Generation","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18750","citing_title":"A Readiness-Driven Runtime for Pipeline-Parallel Training under Runtime Variability","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08016","citing_title":"Video Parallel Scaling: Aggregating Diverse Frame Subsets for VideoLLMs","ref_index":3,"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":6,"is_internal_anchor":false},{"citing_arxiv_id":"2601.21619","citing_title":"On the Overscaling Curse of Parallel Thinking: System Efficacy Contradicts Sample Efficiency","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08696","citing_title":"Structured Recurrent Mixers for Massively Parallelized Sequence Generation","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19341","citing_title":"Evaluation-driven Scaling for Scientific Discovery","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ","json":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ.json","graph_json":"https://pith.science/api/pith-number/XEW3BAY3X5IMYZZ66UOJKNBEXJ/graph.json","events_json":"https://pith.science/api/pith-number/XEW3BAY3X5IMYZZ66UOJKNBEXJ/events.json","paper":"https://pith.science/paper/XEW3BAY3"},"agent_actions":{"view_html":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ","download_json":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ.json","view_paper":"https://pith.science/paper/XEW3BAY3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.10475&json=true","fetch_graph":"https://pith.science/api/pith-number/XEW3BAY3X5IMYZZ66UOJKNBEXJ/graph.json","fetch_events":"https://pith.science/api/pith-number/XEW3BAY3X5IMYZZ66UOJKNBEXJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ/action/storage_attestation","attest_author":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ/action/author_attestation","sign_citation":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ/action/citation_signature","submit_replication":"https://pith.science/pith/XEW3BAY3X5IMYZZ66UOJKNBEXJ/action/replication_record"}},"created_at":"2026-07-05T11:03:38.817883+00:00","updated_at":"2026-07-05T11:03:38.817883+00:00"}