{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:PUSHK26RYUT65IRF2HJ47GK7W2","short_pith_number":"pith:PUSHK26R","schema_version":"1.0","canonical_sha256":"7d24756bd1c527eea225d1d3cf995fb6a5eaea0be5362d70ab1712618bfb7c58","source":{"kind":"arxiv","id":"2303.17564","version":3},"attestation_state":"computed","paper":{"title":"BloombergGPT: A Large Language Model for Finance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"BloombergGPT, a 50 billion parameter model trained on financial plus general data, outperforms prior models on financial tasks while preserving general LLM performance.","cross_cats":["cs.AI","cs.CL","q-fin.GN"],"primary_cat":"cs.LG","authors_text":"David Rosenberg, Gideon Mann, Mark Dredze, Ozan Irsoy, Prabhanjan Kambadur, Sebastian Gehrmann, Shijie Wu, Steven Lu, Vadim Dabravolski","submitted_at":"2023-03-30T17:30:36Z","abstract_excerpt":"The use of NLP in the realm of financial technology is broad and complex, with applications ranging from sentiment analysis and named entity recognition to question answering. Large Language Models (LLMs) have been shown to be effective on a variety of tasks; however, no LLM specialized for the financial domain has been reported in literature. In this work, we present BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific data"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2303.17564","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-30T17:30:36Z","cross_cats_sorted":["cs.AI","cs.CL","q-fin.GN"],"title_canon_sha256":"ce000b4019446a1232badc49fdffe0f3fa25a4751e0d195e87cf1b1f46bd0aff","abstract_canon_sha256":"3c618fff827861fdc6b1d501a9e1e1d2a66362df35cf96dc55522e3df3e43035"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:47:53.394418Z","signature_b64":"xIVEqz99PRJVIkww6RxvXbQKyYUeKtNSnM5iyu+o5xqBDo9JqiA+RiKV3m4TP8grTzasVg+hc1QhFm41KF/4AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7d24756bd1c527eea225d1d3cf995fb6a5eaea0be5362d70ab1712618bfb7c58","last_reissued_at":"2026-05-18T03:47:53.393747Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:47:53.393747Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BloombergGPT: A Large Language Model for Finance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"BloombergGPT, a 50 billion parameter model trained on financial plus general data, outperforms prior models on financial tasks while preserving general LLM performance.","cross_cats":["cs.AI","cs.CL","q-fin.GN"],"primary_cat":"cs.LG","authors_text":"David Rosenberg, Gideon Mann, Mark Dredze, Ozan Irsoy, Prabhanjan Kambadur, Sebastian Gehrmann, Shijie Wu, Steven Lu, Vadim Dabravolski","submitted_at":"2023-03-30T17:30:36Z","abstract_excerpt":"The use of NLP in the realm of financial technology is broad and complex, with applications ranging from sentiment analysis and named entity recognition to question answering. Large Language Models (LLMs) have been shown to be effective on a variety of tasks; however, no LLM specialized for the financial domain has been reported in literature. In this work, we present BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific data"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the internal benchmarks and chosen financial data sources accurately reflect real-world usage and that the performance gains are not due to dataset-specific artifacts or evaluation choices.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"BloombergGPT is a 50B parameter LLM trained on a 708B token mixed financial and general dataset that outperforms prior models on financial benchmarks while preserving general LLM performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"BloombergGPT, a 50 billion parameter model trained on financial plus general data, outperforms prior models on financial tasks while preserving general LLM performance.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a89833c5bc818881b7ff6f8cb2096d8697ff892e4f0cd9653527f9be64826198"},"source":{"id":"2303.17564","kind":"arxiv","version":3},"verdict":{"id":"fb5079f8-d635-4cc5-8715-ec8dd5937eec","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T23:14:46.985302Z","strongest_claim":"Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks.","one_line_summary":"BloombergGPT is a 50B parameter LLM trained on a 708B token mixed financial and general dataset that outperforms prior models on financial benchmarks while preserving general LLM performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the internal benchmarks and chosen financial data sources accurately reflect real-world usage and that the performance gains are not due to dataset-specific artifacts or evaluation choices.","pith_extraction_headline":"BloombergGPT, a 50 billion parameter model trained on financial plus general data, outperforms prior models on financial tasks while preserving general LLM performance."},"references":{"count":140,"sample":[{"doi":"","year":1908,"title":"FinBERT: Financial Sentiment Analysis with Pre-trained Language Models","work_id":"3dd01f6f-6c0f-4a47-a6dd-6ca15bc4e219","ref_index":1,"cited_arxiv_id":"1908.10063","is_internal_anchor":true},{"doi":"","year":2022,"title":"PLATO - XL : Exploring the large-scale pre-training of dialogue generation","work_id":"c1c6d296-00b7-46ca-b0d9-8274a884fa42","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.18653/v1/d19-1371","year":2019,"title":"S ci BERT : A pretrained language model for scientific text","work_id":"f4322284-a0f8-4f17-b855-c0eccacb4545","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pages 610--623","work_id":"717f4c67-3986-48a0-b9f6-50e30b658cbf","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2009,"title":"The fifth PASCAL recognizing textual entailment challenge","work_id":"b6e103a5-a597-4905-b033-4a41b167e7ac","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":140,"snapshot_sha256":"0af8286cfd92006a17cd11e70fb5a030bf5babcb880c2ba0c7e9ded791457812","internal_anchors":32},"formal_canon":{"evidence_count":2,"snapshot_sha256":"22a43aa6f5f72f4089a51719a6fac9a632a9ea4598046a111477a4a1d49cda8f"},"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":"2303.17564","created_at":"2026-05-18T03:47:53.393859+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.17564v3","created_at":"2026-05-18T03:47:53.393859+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.17564","created_at":"2026-05-18T03:47:53.393859+00:00"},{"alias_kind":"pith_short_12","alias_value":"PUSHK26RYUT6","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"PUSHK26RYUT65IRF","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"PUSHK26R","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":83,"internal_anchor_count":83,"sample":[{"citing_arxiv_id":"2607.08346","citing_title":"Grounded Event Extraction from SEC 8-K Filings with a Fine-Grained Taxonomy","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2606.25984","citing_title":"InvestPhilBench: A Multi-Layer Benchmark for Evaluating Large Language Model Procedural Reasoning in Expert Investment Philosophy","ref_index":56,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24950","citing_title":"MacroLens: A Multi-Task Benchmark for Contextual Financial Reasoning under Macroeconomic Scenarios","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2606.26173","citing_title":"AlgoEvolve: LLM-driven Meta-evolution of Algorithmic Trading Programs","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2606.27316","citing_title":"LLM-Based Examination of Eligibility Criteria from Securities Prospectuses at the German Central Bank","ref_index":64,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18192","citing_title":"The Stanford EDGAR Filings Dataset: Reconstructing U.S. Corporate and Financial Disclosures into Layout-Faithful and Token-Efficient Pretraining Data","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2607.01740","citing_title":"Meta-Benchmarks for Financial-Services LLM Evaluation","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2606.13100","citing_title":"LEDGER: A Long-Context Benchmark of Corporate Annual Reports for Grounded Financial Retrieval and Extraction","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2606.10127","citing_title":"Data-Driven Automation","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2606.08283","citing_title":"Macro Economists in the Machine: A Multi-Agent LLM Framework for Commodity-Related ETF Portfolio Construction","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2606.08285","citing_title":"Beyond Agent Architecture: Execution Assumptions and Reproducibility in LLM-Based Trading Systems","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2606.05868","citing_title":"YouZhi: Towards High-Concurrency Financial LLMs via Adaptive GQA-to-MLA Transition","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2607.01023","citing_title":"Evidence-Supported Credit Risk Report Generation Using News-Centric Financial Knowledge Graphs","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2606.05781","citing_title":"Domain-Adapted Small Language Models with Hybrid Post-Processing: Achieving Cost-Efficient, Low-Latency Multi-Label Structured Prediction via LoRA Fine-Tuning on Scarce Data","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2606.05004","citing_title":"SharedRequest: Privacy-Preserving Model-Agnostic Inference for Large Language Models","ref_index":73,"is_internal_anchor":true},{"citing_arxiv_id":"2607.00856","citing_title":"Shapley in Context: Explaining Financial Language with Domain Expertise","ref_index":286,"is_internal_anchor":true},{"citing_arxiv_id":"2606.02530","citing_title":"SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2606.02920","citing_title":"Fast Unlearning at Scale via Margin Self-Correction","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2605.25311","citing_title":"Recursive Multi-Agent Trading System: Iterative Optimized Portfolio Strategy Under Geopolitical Uncertainty","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2606.31608","citing_title":"CLExEval: A Human-in-the-Loop Framework for Qualitative Evaluation of LLM Clinical Reasoning","ref_index":92,"is_internal_anchor":true},{"citing_arxiv_id":"2605.30363","citing_title":"Enhancing Regime Shift Detection Using Unstructured Data: A Study on the Treasury Market","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.23243","citing_title":"Are Frontier LLMs Ready for Cybersecurity? Evidence for Vertical Foundation Models from Dual-Mode Vulnerability Benchmarks","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.24490","citing_title":"Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2605.25030","citing_title":"MimirRAG: A Multi-Agent RAG Framework for Financial Data Retrieval with Metadata Integration","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.25141","citing_title":"LLM Agent Based Renewable Energy Forecasting Using Edge and IoT Data A Review of Solar Wind Weather and Grid Aware Decision Support","ref_index":50,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2","json":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2.json","graph_json":"https://pith.science/api/pith-number/PUSHK26RYUT65IRF2HJ47GK7W2/graph.json","events_json":"https://pith.science/api/pith-number/PUSHK26RYUT65IRF2HJ47GK7W2/events.json","paper":"https://pith.science/paper/PUSHK26R"},"agent_actions":{"view_html":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2","download_json":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2.json","view_paper":"https://pith.science/paper/PUSHK26R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.17564&json=true","fetch_graph":"https://pith.science/api/pith-number/PUSHK26RYUT65IRF2HJ47GK7W2/graph.json","fetch_events":"https://pith.science/api/pith-number/PUSHK26RYUT65IRF2HJ47GK7W2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2/action/storage_attestation","attest_author":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2/action/author_attestation","sign_citation":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2/action/citation_signature","submit_replication":"https://pith.science/pith/PUSHK26RYUT65IRF2HJ47GK7W2/action/replication_record"}},"created_at":"2026-05-18T03:47:53.393859+00:00","updated_at":"2026-05-18T03:47:53.393859+00:00"}