Connect with us

Accounting

Microsoft researchers teach LLMs to use spreadsheets well

Published

on

Large language models like ChatGPT have traditionally had trouble reading and interacting with spreadsheets, limiting their application in this realm, but recent research from Microsoft claims to have found an answer. 

The paper, SPREADSHEETLLM: Encoding Spreadsheets for Large Language Models, described the problems LLMs typically face with spreadsheets and proposed what it called the “SheetCompressor” framework to address them. 

The issue LLMs have with spreadsheets has to do with tokenage requirements. LLMs, generally, run on “tokens,” which are the basic units of data the model processes. Tokens are words, character sets, or combinations of words and punctuation that are used by large language models to decompose text into. LLMs operate by converting input text into a series of tokens, which the model then uses to understand and generate responses. 

The number of tokens determines the computational cost and capacity needed to handle the input, making token management crucial, especially for complex data like spreadsheets. For example, the phrase, “I heard a dog bark loudly at a cat” would be represented by eight tokens, one for each unique word. In order to preserve system resources, many LLMs have token limits, but even in a limitless environment, complex jobs are resource intensive, with significant computational effort that affects both performance and efficiency. 

Typically, each part of a spreadsheet — even blank cells or repeating cells or those with irrelevant information — costs tokens, meaning even a simple spreadsheet has a much higher token requirement than traditional text. Furthermore, LLMs often struggle with spreadsheet-specific features such as cell addresses and formats, complicating their ability to effectively parse and utilize spreadsheet data. These challenges have limited just how much generative AI models can be applied to reading and interacting with spreadsheets. Considering how many spreadsheets the profession tends to use, this consequently limits their application towards deep accounting work. 

What Microsoft researchers discovered, in short, is that the LLM does not need to burn tokens reading and processing the entire spreadsheet. Instead people can create a compressed version of the document to function as something like an index, with markers or “anchors” indicating especially important information like totals. Additional compression comes from grouping together similar types of data like date columns. So, in a sense, the LLM does not work through the spreadsheet itself but instead references it via a much more efficient index. 

Complex spreadsheets are further supported through a concept called “chain of spreadsheet,” which is similar to “chain of thought” prompting. The method unfolds in two stages. First, the model identifies the table that is relevant to the query and determines the precise boundaries of the relevant content. This step ensures that only pertinent data is considered in the subsequent analysis. Then, the query and the identified table section are re-input into the LLM. The model then processes this information to generate an accurate response to the query.

“Through the CoS, SPREADSHEETLLM effectively handles complex spreadsheets by breaking down the process into manageable parts, thus enabling precise and context-aware responses,” said the paper. 

Experiments with this method found that it significantly increased performance on larger spreadsheets where token limits are a particular challenge. The F1 score (which is used to measure the accuracy of an AI model) for massive spreadsheets was 75% higher than GPT-4 and 19% higher than TableSense-CNN, another spreadsheet methodology for AI; for large spreadsheets, the difference was 45% and 17% respectively; for medium spreadsheets it was 13% and 5%; and for small spreadsheets it was 8%. Overall, the results show that while the method gets more effective the larger the spreadsheet, it can still improve the efficiency of even small spreadsheets. 

“Through a novel encoding method, SHEETCOMPRESSOR, this framework effectively addresses the challenges posed by the size, diversity, and complexity inherent in spreadsheets,” the paper concluded. “It achieves a substantial reduction in token usage and computational costs, enabling practical applications on large datasets. The fine-tuning of various cutting-edge LLMs further enhances the performance of spreadsheet understanding. Moreover, Chain of Spreadsheet, the framework’s extension to spreadsheet downstream tasks illustrates its broad applicability and potential to transform spreadsheet data management and analysis, paving the way for more intelligent and efficient user interactions.”

Implications

Donny Shimamoto, founder and managing director of accounting tech-focused accounting firm IntrapriseTechKnowlogies said, by enabling LLMs to “understand” tabular spreadsheets, accountants will have increased ability to either summarize or analyze a set of data. More than that, however, he said this will likely allow even non-accountants to do the same, removing the accountant as the middle person. However while some accountants may see this as a threat, he said what this would mainly do is clear the majority of simple inquiries from their plates, letting them save their energy for more complex questions and deeper analysis.

“Implementing something like this will require good testing to ensure that the risk of hallucinations is minimized, especially if it is going to help provide non-accountants with information to support decision-making,” said Shimamoto.

David Wood, a Bringham Young University accounting professor who specializes in AI within the profession, raised a similar point, as it would allow those without significant technical knowledge to do the same kinds of tasks that, previously, could only be done by seasoned accounting experts. He raised the example of novices being able to use generative AI to make spreadsheets that only expert professionals could put together. However, while he thinks this could be possible soon, he said that, despite the Microsoft research, it hasn’t arrived just yet.

“However, there are at least three challenges holding back using GenAI with spreadsheets: the size and complexity of the spreadsheets, and the required accuracy for most uses of spreadsheets. This paper takes a large step in the right direction, but it doesn’t solve all the challenges and more work will still be needed in each of these three areas. It would be a mistake to assume that after reading this paper, we have fully figured out how to use spreadsheets and GenAI together. More work is still needed. … I think the path these researchers are taking is significant, but the research “hasn’t arrived yet” meaning that more work is needed. The accuracy rates are just not high enough…yet. Hopefully this paves the way for the next researcher to move it forward further.” he said in an email.

Continue Reading

Accounting

GASB issues guidance on capital asset disclosures

Published

on

The Governmental Accounting Standards Board issued guidance today that will require separate disclosures for certain types of capital assets for the purposes of note disclosures.

GASB Statement No. 104, Disclosure of Certain Capital Assets, also establishes requirements and additional disclosures for capital assets held for sale. 

The statement requires certain types of assets to be disclosed separately in the note disclosures about capital assets. The intent is to allow users to make better informed decisions and to evaluate accountability. The requirements are effective for fiscal years beginning after June 15, 2025, and all reporting periods thereafter, though earlier application is encouraged.

The guidance requires separate disclosures for four types of capital assets:

  1. Lease assets reported under Statement 87, by major class of underlying asset;
  2. Intangible right-to-use assets recognized by an operator under Statement 94, by major class of underlying asset;
  3. Subscription assets reported under Statement 96; and,
  4. Intangible assets other than those listed in items 1-3, by major class of asset.

Under the guidance, a capital asset is a capital asset held for sale if the government has decided to pursue the sale of the asset, and it is probable the sale will be finalized within a year of the financial statement date. A government should disclose the historical cost and accumulated depreciation of capital assets held for sale, by major class of asset.

Continue Reading

Accounting

On the move: RRBB hires tax partner

Published

on

Uddin-Suha-RRBB.jpg
Suha Uddin

BRIAN BOUMAN MEMORY CREATIO

Suha Uddin was hired as a tax partner at RRBB Advisors, Somerset. 

Sax, Paterson, announced that its annual run/walk event SAX 4 Miler, supporting the Child Life Department at St. Joseph’s Children’s Hospital in Paterson, has achieved $1 million in total funds raised since its inception in 2012.    

Withum, Princeton, rolled out a new outsourcing service offering as part of its sustainability and ESG practice designed to help companies comply with the European Corporate Sustainability Reporting Directive, the mandate requires reporting of detailed sustainability performance as it pertains to the European Sustainability Reporting Standards , effective January 2023.

Continue Reading

Accounting

Armanino takes on minority investment from Further Global

Published

on

Top 25 Firm Armanino LLP has taken on a strategic minority investment from private equity firm Further Global Capital Management.

The deal, which closed today, is the latest in the series of investments by private equity in large accounting firms that began in 2021 — but with a key difference, Armanino CEO Matt Armanino told Accounting Today.

“What’s maybe the punchline here — what’s really unique, I think — is that we wanted to focus on a minority investment that allowed us to retain not just operational control of the business, but ownership control of the business,” he said. “Those are some of the guiding principles that we’ve been thinking about over the last number of years, and we felt like if we could accomplish those things strategically with the right partner, it would really be just a home run, and that’s where we think we’ve landed.”

As is common with CPA firms taking on private equity investment, Armanino LLP will restructure to an alternative practice structure, splitting into two independently owned and governed professional-services entities: Armanino LLP, a licensed CPA firm wholly owned by individual CPAs, will provide attest services to clients, and Armanino Advisory LLC, a consulting and advisory firm, will perform non-attest services.

Inside the deal

As have many large firms, Armanino LLP had been looking at private equity for some time.

“We’ve been analyzing the PE trend over the last few years and our discussions with Further Global actually began several years ago, and along the way we confirmed our initial inclination that Further Global would be a great partner for us,” CEO Armanino said.

“We had the opportunity to meet with dozens of leading private equity firms,” he explained. “Ultimately we concluded that Further Global would be the best partner for us based on their expertise in partnering with professional service businesses in particular, and our desire for a minority deal structure.”

Matt Armanino
Matt Armanino

Robert Mooring

While citing Further Global’s “deep domain expertise” in financial services and business services firms, Armanino noted that this would be the PE firm’s first foray into the accounting profession: “This is their first accounting firm deal, and I think they’re only focused on this one at this time.”

An employee-owned PE firm, Further Global invests in companies in the business services and financial services industries, and has raised over $2.2 billion of capital.

Guggenheim Securities LLC served as the financial advisor and sole private placement agent to Armanino LLP, while Hunton Andrews Kurth LLP acted as its legal counsel. Further Global was advised by Pointe Advisory, with Kirkland & Ellis as legal counsel.

“Armanino ranks as high as any CPA firm in the country with the private equity community,” commented Allan Koltin, CEO of Koltin Consulting Group, who has advised Armanino for over two decades. “Their deal with Further Global fit just like a glove. They will keep control and now have the capital structure to compete on the biggest of stages.”

Internally, the Armanino partner group was unanimous in its support for the deal — and in its insistence on only selling a minority stake.

“We’ve had transparent discussions at the leadership level around not only adding an outside investor, but we knew very early on that a minority investment was the best path forward for us, and we were very excited that there was unanimous support from the entire partnership group around that decision,” Armanino said. “This structure is also going to allow the long-term owners and partners of Armanino to maintain full control over our day-to-day operations, and the proud culture that we’ve built.”

“No other firm in the Top 25 has a structure like this, and I think that’s pretty significant,” he added.

Capital plans

The goal of the deal is to give Armanino the capital it needs to take itself to a new level of growth while also addressing some of the most pressing challenges in accounting: investing in technology, pursuing inorganic growth through M&A, and attracting and retaining talent.

The firm has always been tech-forward, and recently has been a major pioneer in artificial intelligence.

“The capital will enable us to fast-track our investments in advanced technology solutions, particularly AI,” said Matt Armanino. “We’ve seen growing desire from our clients to deploy real applications for AI solutions. And while we’ve been at the forefront of automation and AI since the early days, with the development of our AI Lab a few years ago, innovative AI-driven solutions that address our clients’ most urgent challenges remain a top priority for us.”

Beyond technology investments, the firm plans to continue its aggressive M&A strategy, which has brought on 19 acquisitions since 2019.

“Those transactions have allowed us to expand our capabilities and enter into new markets and drive greater value to our clients,” said Armanino. “And we think we can accelerate that now with this capital structure that we have.”

All that M&A has brought the firm a lot of fresh talent, but no firm these days has enough, and that’s a third purpose for the new capital.

“We think there remains a lot of ripe talent across the country out there,” he said. “I think the capital will support our efforts to attract, retain, develop and reward top talent by investing in people who drive our entrepreneurial spirit here at the firm.”

The deal will allow the firm to reward top talent, for instance through equity plans that allow them to extend the firm’s ownership culture beyond the partner group that it has traditionally been restricted to.

“In many cases, for our most senior employees today, there’s not a natural mechanism to align their effort to the success of the firm to the growth of our enterprise value and how that ultimately rewards them,” explained Armanino. “And we are very excited that we have new mechanisms, and plans in place, that are going to allow us to do that very well, and effectively push down the benefits of ownership and that ownership culture to our most senior employees.”

“Finally,” he added, “speaking to our innovative culture — and that’s a big part of our brand — the capital will empower us to say ‘Yes’ more frequently to great ideas, to entrepreneurial ideas and initiatives that truly make a difference for our clients and set us apart as a leader in this industry.”

Continue Reading

Trending