Pre_Trained LLM

This dataset was created to provide comparative statistics on various general-purpose and pre-trained financial large language models (LLMs) based on their financial capabilities. We also emphasize on which specific characteristics contributed to their financial capacity. We also provide data on whether the models were open- or closed-sourced. Our intent was to generate a categorical heatmap from the CSV data. The listed LLMs possess wide range of financial task capabilities which includes Real-time forecasting, Quantitative trading, Insolvency forecasting, Mergers and Acquisition Forecasting, Stock Prediction, News Headlines, Named Entity Recognition, Relation Extraction, Causal Classification, Sentiment Analysis, Financial News Multi-Label Classification, Financial Question Answering, Structured Financial Reasoning, Credit Scoring, Fraud Detection, Portfolio Optimization, Financial Reporting Analysis. The specialty of these models often stems from their training data, efficiency, or architectural adaptations:\u2022 Data and Specialization: Models like BloombergGPT were few-shot trained and used 9 specific finance datasets. Mixed data approach which provided data quality and specificity was key. FinBERT uses automated labeling and reinforcement learning of stock prices to improve prediction.\u2022 Efficiency and Cost: FinGPT and Instruct FinGPT are highlighted as cost-effective options. FinGPT improves financial reasoning while demonstrating cost-effectiveness by leveraging lightweight adaptation methods such as LoRA and QLoRA. Instruct FinGPT is trained on LLaMA 7B and is noted as costing less than $300.\u2022 Architectural Features: GPT4 emphasizes learning cross-sequence patterns, multi-modal learning, and handling vast amounts of data. Models like FLANG-BERT and FLANG-ELECTRA are domain-independent LLMs that utilize differential masked 8200 words and phrases.\u2022 Multi-modal Integration: FinTral is distinguished by its capability for multi-modal data processing, specializing in financial understanding, real-time analysis, and decision-making tool integration for quantitative tasks.The key idea to take from the categorical heatmap is to identify and acknowledge the huge variety of Financial Language Models that are present in the industry.