SoftBank Group reported a 124% jump in quarterly profit on resilient AI demand that's supporting startup valuations and chip unit sales, a boost to its aggressive data center investment plans. The Tokyo-based company reported net income of 517.18 billion ( 3.5 billion) in its fiscal fourth quarter. It was helped by the Vision Fund, which swung to a profit of 26.1 billion. The earnings come at a critical juncture for SoftBank as it plans to invest 30 billion in OpenAI while leading a 100 billion foray into building AI hardware in the U.S. Maintaining a healthy cash flow and balance sheet is key to securing the billions of dollars needed at minimum cost.

Meta is set to report its first quarter earnings on Wednesday after the bell, and investors will be looking for news on whether the company met its quarterly revenue goals of somewhere between 39.5bn and 41.8bn. Wall Street is projecting the company will post 41.36bn in revenue on 5.21 in earnings per share. While Meta has repeatedly beaten Wall Street expectations in the past few quarters, analysts were disappointed by the first quarter revenue outlook Meta chief executive Mark Zuckerberg shared at the end of 2024. The company is also planning on spending up to 65bn on AI infrastructure by the end of 2025. Uncertainty over Donald Trump's sweeping tariffs may yet roil ad markets, clouding the company's financial outlook for near future quarters.

Tesla brought in 20 percent less automotive revenue at the end of last year compared to the year previous, the company reported today, as demand for its electric cars appear to have dipped precipitously across the globe. The drop exceeded even some pessimistic Wall Street analysts' predictions. By late afternoon, before CEO Elon Musk and other company leaders appeared for a quarterly update call for investors, stock prices appeared relatively stable on the news. Overall, however, the electric automaker's stock price is down more than 40 percent from its late 2024 high. In a slide deck prepared for investors, Tesla pinned the drop on declines in deliveries, some which it said were related to the need to retool some of its production lines for modified versions of its best-selling electric cars.

Hugging Face, a company that hosts open source artificial intelligence models and software, announced today that it has acquired Pollen Robotics, the French startup behind the bug-eyed, two-armed, humanoid robot called Reachy 2. Hugging Face plans to sell the robot and will also allow developers to download, modify, and suggest improvements to its code. "It's really important for robotics to be as open source as possible," says Clément Delangue, chief executive of Hugging Face. "When you think about physical objects doing physical things at work and at home, the level of trust and transparency I need is much higher than for something I chat with on my laptop." Simon Alibert and Rémi Cadene are research engineers in AI and robotics at Hugging Face. In videos shared by Pollen Robotics, Reachy 2 can be seen performing tricks like tidying coffee mugs and picking up fruit.

OpenAI and its CEO Sam Altman have counter-sued its co-founder turned competitor Elon Musk, accusing the billionaire of unfair and fraudulent business practices. Specifically, ChatGPT's owners claim that Musk's 97.375 billion offer to buy it out in February was a "sham bid" deliberately intended to impede OpenAI's efforts to raise funding. "Through press attacks, malicious campaigns broadcast to Musk's more than 200 million followers on the social media platform he controls [X], a pretextual demand for corporate records, harassing legal claims, and a sham bid for OpenAI's assets, Musk has tried every tool available to harm OpenAI," read the lawsuit. Filed to a California district court on Wednesday, OpenAI's countersuit alleges that Musk's offer to purchase the AI organisation for 97.375 billion was not genuine, and was in fact orchestrated to gain an unfair business advantage. Though Musk was one of OpenAI's founders, he has since left and founded competitor xAI.

Many of the recent breakthroughs in language modeling have resulted from scaling effectively the same model architecture to larger datasets. In this vein, recent work has highlighted performance gains from increasing training dataset size and quality, suggesting a need for novel sources of large-scale datasets. In this work, we introduce BeanCounter, a public dataset consisting of more than 159B tokens extracted from businesses' disclosures. We show that this data is indeed novel: less than 0.1% of BeanCounter appears in Common Crawl-based datasets and the data is an order of magnitude larger than datasets relying on similar sources. Given the data's provenance, we hypothesize that BeanCounter is comparatively more factual and less toxic than web-based datasets. Exploring this hypothesis, we find that many demographic identities occur with similar prevalence in BeanCounter but with significantly less toxic context relative to other datasets. To demonstrate the utility of BeanCounter, we evaluate and compare two LLMs continually pre-trained on BeanCounter with their base models. We find an 18-33% reduction in toxic generation and improved performance within the finance domain for the continually pretrained models. Collectively, our work suggests that BeanCounter is a novel source of low-toxicity and high-quality domain-specific data with sufficient scale to train multi-billion parameter LLMs.

Fact-checking is extensively studied in the context of misinformation and disinformation, addressing objective inaccuracies. However, a softer form of misinformation involves responses that are factually correct but lack certain features such as clarity and relevance. This challenge is prevalent in formal Question-Answer (QA) settings such as press conferences in finance, politics, sports, and other domains, where subjective answers can obscure transparency. Despite this, there is a lack of manually annotated datasets for subjective features across multiple dimensions. To address this gap, we introduce SubjECTive-QA, a human annotated dataset on Earnings Call Transcripts' (ECTs) QA sessions as the answers given by company representatives are often open to subjective interpretations and scrutiny. The dataset includes 49, 446 annotations for long-form QA pairs across six features: Assertive, Cautious, Optimistic, Specific, Clear, and Relevant. These features are carefully selected to encompass the key attributes that reflect the tone of the answers provided during QA sessions across different domains. Our findings are that the best-performing Pre-trained Language Model (PLM), RoBERTa-base, has similar weighted F1 scores to Llama-3-70b-Chat on features with lower subjectivity, such as Relevant and Clear, with a mean difference of 2.17% in their weighted F1 scores. The models perform significantly better on features with higher subjectivity, such as Specific and Assertive, with a mean difference of 10.01% in their weighted F1 scores.

This study provides the first comprehensive assessment of consistency and reproducibility in Large Language Model (LLM) outputs in finance and accounting research. We evaluate how consistently LLMs produce outputs given identical inputs through extensive experimentation with 50 independent runs across five common tasks: classification, sentiment analysis, summarization, text generation, and prediction. Using three OpenAI models (GPT-3.5-turbo, GPT-4o-mini, and GPT-4o), we generate over 3.4 million outputs from diverse financial source texts and data, covering MD&As, FOMC statements, finance news articles, earnings call transcripts, and financial statements. Our findings reveal substantial but task-dependent consistency, with binary classification and sentiment analysis achieving near-perfect reproducibility, while complex tasks show greater variability. More advanced models do not consistently demonstrate better consistency and reproducibility, with task-specific patterns emerging. LLMs significantly outperform expert human annotators in consistency and maintain high agreement even where human experts significantly disagree. We further find that simple aggregation strategies across 3-5 runs dramatically improve consistency. Simulation analysis reveals that despite measurable inconsistency in LLM outputs, downstream statistical inferences remain remarkably robust. These findings address concerns about what we term "G-hacking," the selective reporting of favorable outcomes from multiple Generative AI runs, by demonstrating that such risks are relatively low for finance and accounting tasks.