Only two of seven LLMs produce positive returns on live Polymarket data, with MiMo-V2-Flash at 17.6% CWR and Gemini-3-Flash at 6.2% CWR while the other five lose money.
hub Canonical reference
BloombergGPT: A Large Language Model for Finance
Canonical reference. 100% of citing Pith papers cite this work as background.
80
Pith papers citing it
Background
100% of classified citations
abstract
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 dataset yet, augmented with 345 billion tokens from general purpose datasets. We validate BloombergGPT on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect our intended usage. 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. Additionally, we explain our modeling choices, training process, and evaluation methodology. We release Training Chronicles (Appendix C) detailing our experience in training BloombergGPT.
hub tools
citation-role summary
background 11
citation-polarity summary
roles
background 11polarities
background 11representative citing papers
First study of 1,899 MCP servers finds eight distinct vulnerabilities (only three traditional), 7.2% with general issues, 5.5% with tool poisoning, and 66% with code smells, urging MCP-specific security practices.
A meta-benchmarking framework organizes 452 LLM benchmarks into 41 O*NET Generalized Work Activities and 38 BIAN domains, using discrimination-coverage-recency weights to scale K-factors in an Elo tournament for comparable financial-services scores.
Multi-agent LLM system Agora under Sealed Joint Search conditions produces +1.87 holdout Sharpe on CSI 1000 over a 91-day sealed period, exceeding the best baseline at +1.334 under favorable seed.
LLMs are applied in a generative pipeline for extracting, normalizing, and interpreting eligibility criteria from securities prospectuses, achieving up to 91% precision in document-level decisions with a conservative bias.
Composite LLM scoring saturates on expert investment frameworks while Gate Reconstruction Accuracy still reveals a clear procedural deficit at the frontier.
MacroLens is a point-in-time multi-signal benchmark dataset and seven tasks for evaluating contextual financial reasoning models under macroeconomic scenarios.
SEFD reconstructs SEC filings into MultiMarkdown to create a 152B-token financial pretraining corpus with low overlap to existing data and introduces EDGAR-Forecast and EDGAR-OCR benchmarks.
LEDGER provides a corpus of 4,999 annual reports with 31 labeled KPIs and three benchmarks for page-level retrieval, needle-in-haystack lookup, and full KPI extraction from long documents.
MUSE framework shows LLM conformity to user pushback arises from both sycophantic alignment and epistemic uncertainty, with both increasing when users appear expert or suggestions seem plausible.
QuantEvolver applies reinforcement fine-tuning to evolve an LLM policy for generating executable alpha factor expressions, yielding higher-quality and more complementary factors than prompt-based baselines on market benchmarks.
AutoRedTrader generates synthetic financial misinformation via behavioral bias manipulation and agent feedback to red-team LLM trading agents, reaching 69% exposure and 26.67% attack success on Bitcoin data simulations.
Constrained LLM agents discover cryptocurrency factors that produce a portfolio with 44.55% annualized return and Sharpe ratio of 1.55 in pure out-of-sample 2024-2026 testing after trading costs.
AWASH detects AI-washing via cross-modal inconsistency reasoning on a new trimodal benchmark of 88k corporate disclosure triplets, achieving F1 0.882 with a CMID network that grounds claims against patents and hiring data.
SynBench benchmarks DP text generators across nine datasets and uses a new MIA to show that public pre-training on portions of private data overestimates synthetic text quality and breaks DP privacy bounds.
Schema-constrained, quote-grounded LLM extraction plus a second-pass quality score yields 601k auditable 8-K event tags whose precision and market reactions both improve with the score.
CLExEval introduces a human-annotated evaluation framework on 40 rare cases that identifies verbosity bias, hidden knowledge paradox, and 68.6% reasoning-to-output mismatch in LLMs while showing LLM-as-a-Judge overestimates reliability.
MASC achieves competitive forget-retain trade-offs in language model unlearning at lower computational cost via margin self-correction and an online stopping criterion on TOFU, MUSE News, and MUSE Books.
SafeSteer restricts reverse KL penalty to safety tokens selected via activation steering, achieving strong safety on seven benchmarks with minimal degradation on five capability benchmarks using only 100 harmful samples and no general data.
IPO-Mine releases a toolkit and large multimodal dataset for structured analysis of IPO filings and shows state-of-the-art models diverge from human judgments on chart quality and misleadingness.
Using FOMC minutes to propose regime-shift candidates and a lenient text check to ratify data-detected candidates, the pipeline reaches F1=0.82 on 26 monetary-policy anchors, beating every data-only baseline.
Distinguishable Deletion unifies knowledge erasure and refusal for LLM unlearning via an energy index that enforces boundaries during training and enables refusal at inference.
A dataset-agnostic framework converts text tool-calling benchmarks to paired audio evaluations via TTS, speaker variation and noise, then evaluates seven omni-modal models showing model- and task-dependent performance with small text-to-voice gaps.
Stateful sessions with incremental KV cache and flash queries allow O(|q|) latency in streaming transformer inference, delivering up to 5.9x speedup over conventional engines while preserving full attention.
citing papers explorer
-
PolyBench: Benchmarking LLM Forecasting and Trading Capabilities on Live Prediction Market Data
Only two of seven LLMs produce positive returns on live Polymarket data, with MiMo-V2-Flash at 17.6% CWR and Gemini-3-Flash at 6.2% CWR while the other five lose money.
-
Model Context Protocol (MCP) at First Glance: Studying the Security and Maintainability of MCP Servers
First study of 1,899 MCP servers finds eight distinct vulnerabilities (only three traditional), 7.2% with general issues, 5.5% with tool poisoning, and 66% with code smells, urging MCP-specific security practices.
-
Meta-Benchmarks for Financial-Services LLM Evaluation
A meta-benchmarking framework organizes 452 LLM benchmarks into 41 O*NET Generalized Work Activities and 38 BIAN domains, using discrimination-coverage-recency weights to scale K-factors in an Elo tournament for comparable financial-services scores.
-
AI Trading's Alpha Singularity: Emergent Market Reasoning through Agent-to-Agent Self-Evolution
Multi-agent LLM system Agora under Sealed Joint Search conditions produces +1.87 holdout Sharpe on CSI 1000 over a 91-day sealed period, exceeding the best baseline at +1.334 under favorable seed.
-
LLM-Based Examination of Eligibility Criteria from Securities Prospectuses at the German Central Bank
LLMs are applied in a generative pipeline for extracting, normalizing, and interpreting eligibility criteria from securities prospectuses, achieving up to 91% precision in document-level decisions with a conservative bias.
-
InvestPhilBench: A Multi-Layer Benchmark for Evaluating Large Language Model Procedural Reasoning in Expert Investment Philosophy
Composite LLM scoring saturates on expert investment frameworks while Gate Reconstruction Accuracy still reveals a clear procedural deficit at the frontier.
-
MacroLens: A Multi-Task Benchmark for Contextual Financial Reasoning under Macroeconomic Scenarios
MacroLens is a point-in-time multi-signal benchmark dataset and seven tasks for evaluating contextual financial reasoning models under macroeconomic scenarios.
-
The Stanford EDGAR Filings Dataset: Reconstructing U.S. Corporate and Financial Disclosures into Layout-Faithful and Token-Efficient Pretraining Data
SEFD reconstructs SEC filings into MultiMarkdown to create a 152B-token financial pretraining corpus with low overlap to existing data and introduces EDGAR-Forecast and EDGAR-OCR benchmarks.
-
LEDGER: A Long-Context Benchmark of Corporate Annual Reports for Grounded Financial Retrieval and Extraction
LEDGER provides a corpus of 4,999 annual reports with 31 labeled KPIs and three benchmarks for page-level retrieval, needle-in-haystack lookup, and full KPI extraction from long documents.
-
It's Not Always Sycophancy: Measuring LLM Conformity as a Function of Epistemic Uncertainty
MUSE framework shows LLM conformity to user pushback arises from both sycophantic alignment and epistemic uncertainty, with both increasing when users appear expert or suggestions seem plausible.
-
From Feedback Loops to Policy Updates: Reinforcement Fine-Tuning for LLM-Based Alpha Factor Discovery
QuantEvolver applies reinforcement fine-tuning to evolve an LLM policy for generating executable alpha factor expressions, yielding higher-quality and more complementary factors than prompt-based baselines on market benchmarks.
-
AutoRedTrader: Autonomous Red Teaming of Trading Agents through Synthetic Misinformation Injection
AutoRedTrader generates synthetic financial misinformation via behavioral bias manipulation and agent feedback to red-team LLM trading agents, reaching 69% exposure and 26.67% attack success on Bitcoin data simulations.
-
From Hypotheses to Factors: Constrained LLM Agents in Cryptocurrency Markets
Constrained LLM agents discover cryptocurrency factors that produce a portfolio with 44.55% annualized return and Sharpe ratio of 1.55 in pure out-of-sample 2024-2026 testing after trading costs.
-
Detecting Corporate AI-Washing via Cross-Modal Semantic Inconsistency Learning
AWASH detects AI-washing via cross-modal inconsistency reasoning on a new trimodal benchmark of 88k corporate disclosure triplets, achieving F1 0.882 with a CMID network that grounds claims against patents and hiring data.
-
SynBench: A Benchmark for Differentially Private Text Generation
SynBench benchmarks DP text generators across nine datasets and uses a new MIA to show that public pre-training on portions of private data overestimates synthetic text quality and breaks DP privacy bounds.
-
Grounded Event Extraction from SEC 8-K Filings with a Fine-Grained Taxonomy
Schema-constrained, quote-grounded LLM extraction plus a second-pass quality score yields 601k auditable 8-K event tags whose precision and market reactions both improve with the score.
-
CLExEval: A Human-in-the-Loop Framework for Qualitative Evaluation of LLM Clinical Reasoning
CLExEval introduces a human-annotated evaluation framework on 40 rare cases that identifies verbosity bias, hidden knowledge paradox, and 68.6% reasoning-to-output mismatch in LLMs while showing LLM-as-a-Judge overestimates reliability.
-
Fast Unlearning at Scale via Margin Self-Correction
MASC achieves competitive forget-retain trade-offs in language model unlearning at lower computational cost via margin self-correction and an online stopping criterion on TOFU, MUSE News, and MUSE Books.
-
SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
SafeSteer restricts reverse KL penalty to safety tokens selected via activation steering, achieving strong safety on seven benchmarks with minimal degradation on five capability benchmarks using only 100 harmful samples and no general data.
-
IPO-Mine: A Toolkit and Dataset for Section-Structured Analysis of Long, Multimodal IPO Documents
IPO-Mine releases a toolkit and large multimodal dataset for structured analysis of IPO filings and shows state-of-the-art models diverge from human judgments on chart quality and misleadingness.
-
Enhancing Regime Shift Detection Using Unstructured Data: A Study on the Treasury Market
Using FOMC minutes to propose regime-shift candidates and a lenient text check to ratify data-detected candidates, the pipeline reaches F1=0.82 on 26 monetary-policy anchors, beating every data-only baseline.
-
Distinguishable Deletion: Unifying Knowledge Erasure and Refusal for Large Language Model Unlearning
Distinguishable Deletion unifies knowledge erasure and refusal for LLM unlearning via an energy index that enforces boundaries during training and enables refusal at inference.
-
From Text to Voice: A Reproducible and Verifiable Framework for Evaluating Tool Calling LLM Agents
A dataset-agnostic framework converts text tool-calling benchmarks to paired audio evaluations via TTS, speaker variation and noise, then evaluates seven omni-modal models showing model- and task-dependent performance with small text-to-voice gaps.
-
Attention Once Is All You Need: Efficient Streaming Inference with Stateful Transformers
Stateful sessions with incremental KV cache and flash queries allow O(|q|) latency in streaming transformer inference, delivering up to 5.9x speedup over conventional engines while preserving full attention.
-
Effective Performance Measurement: Challenges and Opportunities in KPI Extraction from Earnings Calls
Encoder models trained on SEC filings struggle with earnings calls due to domain shift, while LLMs enable open-ended KPI extraction with 79.7% human-verified precision on newly introduced benchmarks.
-
RouteNLP: Closed-Loop LLM Routing with Conformal Cascading and Distillation Co-Optimization
RouteNLP is a closed-loop LLM routing framework using conformal cascading and distillation co-optimization that cut inference costs by 58% in an 8-week enterprise deployment while preserving 91% acceptance and high quality on benchmarks.
-
Cross-Stock Predictability via LLM-Augmented Semantic Networks
LLM filtering of embedding-based stock networks raises long-short Sharpe ratio from 0.742 to 0.820 and cuts max drawdown from -10.47% to -7.85% in 2011-2019 S&P 500 backtests.
-
QRAFTI: An Agentic Framework for Empirical Research in Quantitative Finance
QRAFTI is a multi-agent framework using tool-calling and reflection-based planning to emulate quant research tasks like factor replication and signal testing on financial data.
-
MFMDQwen: Multilingual Financial Misinformation Detection Based on Large Language Model
MFMDQwen is the first open-source LLM for multilingual financial misinformation detection, backed by a new instruction dataset and benchmark on which it outperforms other open-source models.
-
SenseAI: A Human-in-the-Loop Dataset for RLHF-Aligned Financial Sentiment Reasoning
SenseAI is a human-in-the-loop financial sentiment dataset with reasoning processes and market outcomes that reveals predictable LLM error patterns like Latent Reasoning Drift for RLHF alignment.
-
SysTradeBench: An Iterative Build-Test-Patch Benchmark for Strategy-to-Code Trading Systems with Drift-Aware Diagnostics
SysTradeBench evaluates 17 LLMs on 12 trading strategies, finding over 91.7% code validity but rapid convergence in iterative fixes and a continued need for human oversight on critical strategies.
-
PolySwarm: A Multi-Agent Large Language Model Framework for Prediction Market Trading and Latency Arbitrage
PolySwarm aggregates predictions from 50 LLM personas for Polymarket trading using Bayesian combination and divergence metrics, outperforming single models in calibration while adding latency arbitrage via CEX price models.
-
CGCMA: Conditionally-Gated Cross-Modal Attention for Event-Conditioned Asynchronous Fusion
CGCMA separates text-conditioned grounding from lag-aware trust gating to fuse asynchronous price and web data, yielding the highest Sharpe ratio of +0.449 on a new crypto news corpus.
-
Sell Me This Stock: Unsafe Recommendation Drift in LLM Agents
Under manipulated tool outputs, eight LLM agents produce risk-mismatched stock recommendations in 65–99% of turns while NDCG-style quality scores stay nearly identical to clean runs.
-
Understanding Structured Financial Data with LLMs: A Case Study on Fraud Detection
FinFRE-RAG combines importance-guided feature reduction with label-aware retrieval-augmented generation to boost LLM performance on tabular fraud detection across four public datasets while providing human-readable rationales.
-
MM-Telco: Benchmarks and Multimodal Large Language Models for Telecom Applications
MM-Telco creates multimodal benchmarks for telecom and demonstrates that fine-tuned LLMs and VLMs achieve significant performance gains on domain-specific tasks.
-
Learning to Conceal Risk: Controllable Multi-turn Red Teaming for LLMs in the Financial Domain
CoRT achieves 95% average attack success rate on nine LLMs by using iterative risk-concealing prompts and a controller that scores concealment levels on a new 522-instruction financial risk benchmark.
-
The Dark Side of LLMs: Agent-based Attack Vectors for System-level Compromise
Testing 18 LLMs found 94.4% vulnerable to direct prompt injection for malware installation, 83.3% to RAG backdoor attacks, and 100% to inter-agent trust exploitation in multi-agent systems.
-
From Time Series Analysis to Question Answering: A Survey in the LLM Era
A survey proposing a taxonomy of Injective, Bridging, and Internal Alignment paradigms to evolve TSA into user-driven Time Series Question Answering with LLMs.
-
ConfusionPrompt: Practical Private Inference for Online Large Language Models
ConfusionPrompt enables private black-box LLM inference via prompt decomposition and pseudo-prompt mixing, claiming better privacy-utility trade-off than perturbation methods and lower memory use than open-source local models.
-
Jailbreaking Black Box Large Language Models in Twenty Queries
PAIR uses an attacker LLM to iteratively craft effective jailbreak prompts for black-box target LLMs in fewer than 20 queries.
-
SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks
SmoothLLM mitigates jailbreaking attacks on LLMs by randomly perturbing multiple copies of a prompt at the character level and aggregating the outputs to detect adversarial inputs.
-
Evidence-Supported Credit Risk Report Generation Using News-Centric Financial Knowledge Graphs
FinKG-News constructs news-centric financial knowledge graphs to support in-context learning for credit risk report generation across three dimensions, claiming 19-34% quality gains and fewer hallucinations than baselines.
-
AlgoEvolve: LLM-driven Meta-evolution of Algorithmic Trading Programs
An LLM-driven evolutionary framework generates executable trading strategies as Python code and uses a meta-loop to evolve the prompts that guide synthesis.
-
Data-Driven Automation
Dynamic model of data-driven automation with heterogeneous accumulating data and spillovers derives conditions for partial versus full automation, shows asymptotic power-law decay in labor share, generic inefficiency, and with endogenous capital, explosive growth but stagnant long-run wages.
-
YouZhi: Towards High-Concurrency Financial LLMs via Adaptive GQA-to-MLA Transition
YouZhi-LLM applies a layer-adaptive GQA-to-MLA transition plus Ascend-specific distillation and fine-tuning to reduce KV-cache size, yielding up to 2.69× higher concurrency and modest gains on financial benchmarks versus base models.
-
SharedRequest: Privacy-Preserving Model-Agnostic Inference for Large Language Models
SharedRequest is a model-agnostic batch-level framework that mixes prompts with noise and groups equivalent instructions to achieve higher utility and lower query cost than individual differential privacy methods for LLM inference.
-
Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems
MRC computes coalition Shapley credits from performance histories to weight three LLM agents, stabilized by Bayesian mixture and regime multipliers, achieving SR 1.51 and 440.1% cumulative return over 1037 days on 13 crypto assets.
-
Are Frontier LLMs Ready for Cybersecurity? Evidence for Vertical Foundation Models from Dual-Mode Vulnerability Benchmarks
Frontier LLMs exhibit 10-50% false positives in white-box vulnerability detection and 4-8% ground-truth coverage in black-box web testing, with domain-specialized agents and models outperforming them and supporting the case for vertical foundation models in cybersecurity.
-
Reasoning through Verifiable Forecast Actions: Consistency-Grounded RL for Financial LLMs
StockR1 unifies LLM-based financial reasoning and time-series forecasting by emitting verifiable forecast actions that condition a decoder, optimized via consistency-grounded RL to improve accuracy on QA and prediction tasks.