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

New IRS regs put some partnership transactions under spotlight

Published

on

Final regulations now identify certain partnership related-party “basis shifting” transactions as “transactions of interest” subject to the rules for reportable transactions.

The final regs apply to related partners and partnerships that participated in the identified transactions through distributions of partnership property or the transfer of an interest in the partnership by a related partner to a related transferee. Affected taxpayers and their material advisors are subject to the disclosure requirements for reportable transactions. 

During the proposal process, the Treasury and the Internal Revenue Service received comments that the final regulations should avoid unnecessary burdens for small, family-run businesses, limit retroactive reporting, provide more time for reporting and differentiate publicly traded partnerships, among other suggested changes now reflected in the regs.

  • Increased dollar threshold for basis increase in a TOI. The threshold amount for a basis increase in a TOI has been increased from $5 million to $25 million for tax years before 2025 and $10 million for tax years after. 
  • Limited retroactive reporting for open tax years. Reporting has been limited for open tax years to those that fall within a six-year lookback window. The six-year lookback is the 72-month period before the first month of a taxpayer’s most recent tax year that began before the publication of the final regulations (slated for Jan. 14 in the Federal Register). Also, the threshold amount for a basis increase in a TOI during the six-year lookback is $25 million. 
  • Additional time for reporting. Taxpayers have an additional 90 days from the final regulation’s publication to file disclosure statements for TOIs in open tax years for which a return has already been filed and that fall within the six-year lookback. Material advisors have an additional 90 days to file their disclosure statements for tax statements made before the final regulations. 
  • Publicly traded partnerships. Because PTPs are typically owned by a large number of unrelated owners, the final regulations exclude many owners of PTPs from the disclosure rules. 

The identified transactions generally result from either a tax-free distribution of partnership property to a partner that is related to one or more partners of the partnership, or the tax-free transfer of a partnership interest by a related partner to a related transferee.

IRS headquarters

Bloomberg via Getty Images

The tax-free distribution or transfer generates an increase to the basis of the distributed property or partnership property of $10 million or more ($25 million or more in the case of a TOI undertaken in a tax year before 2025) under the rules of IRC Sections 732(b) or (d), 734(b) or 743(b), but for which no corresponding tax is paid. 

The basis increase to the distributed or partnership property allows the related parties to decrease taxable income through increased cost recovery allowances or decrease taxable gain (or increase taxable loss) on the disposition of the property.

Continue Reading

Accounting

Treasury, IRS propose rules on commercial clean vehicles, issue guidance on clean fuels

Published

on

The Treasury Department and the Internal Revenue Service proposed new rules for the tax credit for qualified commercial clean vehicles, along with guidance on claiming tax credits for clean fuel under the Inflation Reduction Act.

The Notice of Proposed Rulemaking on the credit for qualified commercial clean vehicles (under Section 45W of the Tax Code) says the credit can be claimed by purchasing and placing in service qualified commercial clean vehicles, including certain battery electric vehicles, plug-in hybrid EVs, fuel cell electric vehicles and plug-in hybrid fuel cell electric vehicles.  

The credit is the lesser amount of either 30% of the vehicle’s basis (15% for plug-in hybrid EVs) or the vehicle’s incremental cost in excess of a vehicle comparable in size or use powered solely by gasoline or diesel. A credit up to $7,500 can be claimed for a single qualified commercial clean vehicle for cars and light-duty trucks (with a Gross Vehicle Weight Rating of less than 14,000 pounds), or otherwise $40,000 for vehicles like electric buses and semi-trucks (with a GVWR equal to or greater than 14,000 pounds).

“The release of Treasury’s proposed rules for the commercial clean vehicle credit marks an important step forward in the Biden-Harris Administration’s work to lower transportation costs and strengthen U.S. energy security,” said U.S. Deputy Secretary of the Treasury Wally Adeyemo in a statement Friday. “Today’s guidance will provide the clarity and certainty needed to grow investment in clean vehicle manufacturing.”

The NPRM issued today proposes rules to implement the 45W credit, including proposing various pathways for taxpayers to determine the incremental cost of a qualifying commercial clean vehicle for purposes of calculating the amount of 45W credit. For example, the NPRM proposes that taxpayers can continue to use the incremental cost safe harbors such as those set out in Notice 2023-9 and Notice 2024-5, may rely on a manufacturer’s written cost determination to determine the incremental cost of a qualifying commercial clean vehicle, or may calculate the incremental cost of a qualifying clean vehicle versus an internal combustion engine (ICE) vehicle based on the differing costs of the vehicle powertrains.

The NPRM also proposes rules regarding the types of vehicles that qualify for the credit and aligns certain definitional concepts with those applicable to the 30D and 25E credits. In addition, the NPRM proposes that vehicles are only eligible if they are used 100% for trade or business, excepting de minimis personal use, and that the 45W credit is disallowed for qualified commercial clean vehicles that were previously allowed a clean vehicle credit under 30D or 45W. 

The notice asks for comments over the next 60 days on the proposed regulations such as issues related to off-road mobile machinery, including approaches that might be adopted in applying the definition of mobile machinery to off-road vehicles and whether to create a product identification number system for such machinery in order to comply with statutory requirements. A public hearing is scheduled for April 28, 2025.

Clean Fuels Production Credit

The Treasury the IRS also released guidance Friday on the Clean Fuels Production Credit under Section 45Z of the Tax Code.

Section 45Z provides a tax credit for the production of transportation fuels with lifecycle greenhouse gas emissions below certain levels. The credit is in effect in 2025 and is for sustainable aviation fuel and non-SAF transportation fuels.

The guidance includes both a notice of intent to propose regulations on the Section 45Z credit and a notice providing the annual emissions rate table for Section 45Z, which refers taxpayers to the appropriate methodologies for determining the lifecycle GHG emissions of their fuel. In conjunction with the guidance released Friday, the Department of Energy plans to release the 45ZCF-GREET model for use in determining emissions rates for 45Z in the coming days.

“This guidance will help put America on the cutting-edge of future innovation in aviation and renewable fuel while also lowering transportation costs for consumers,” said Adeyemo in a statement. “Decarbonizing transportation and lowering costs is a win-win for America.”

Section 45Z provides a per-gallon (or gallon-equivalent) tax credit for producers of clean transportation fuels based on the carbon intensity of production. It consolidates and replaces pre-Inflation Reduction Act credits for biodiesel, renewable diesel, and alternative fuels, and an IRA credit for sustainable aviation fuel. Like several other IRA credits, Section 45Z requires the Treasury to establish rules for measuring carbon intensity of production, based on the Clean Air Act’s definition of “lifecycle greenhouse gas emissions.”

The guidance offers more clarity on various issues, including which entities and fuels are eligible for the credit, and how taxpayers determine lifecycle emissions. Specifically, the guidance outlines the Treasury and the IRS’s intent to define key concepts and provide certain rules in a future rulemaking, including clarifying who is eligible for a credit.

The Treasury and the IRS intend to provide that the producer of the eligible clean fuel is eligible to claim the 45Z credit. In keeping with the statute, compressors and blenders of fuel would not be eligible.

Under Section 45Z, a fuel must be “suitable for use” as a transportation fuel. The Treasury and the IRS intend to propose that 45Z-creditable transportation fuel must itself (or when blended into a fuel mixture) have either practical or commercial fitness for use as a fuel in a highway vehicle or aircraft. The guidance clarifies that marine fuels that are otherwise suitable for use in highway vehicles or aircraft, such as marine diesel and methanol, are also 45Z eligible.

Specifically, this would mean that neat SAF that is blended into a fuel mixture that has practical or commercial fitness for use as a fuel would be creditable. Additionally, natural gas alternatives such as renewable natural gas would be suitable for use if produced in a manner such that if it were further compressed it could be used as a transportation fuel.

Today’s guidance publishes the annual emissions rate table that directs taxpayers to the appropriate methodologies for calculating carbon intensities for types and categories of 45Z-eligible fuels.

The table directs taxpayers to use the 45ZCF-GREET model to determine the emissions rate of non-SAF transportation fuel, and either the 45ZCF-GREET model or methodologies from the International Civil Aviation Organization (“CORSIA Default” or “CORSIA Actual”) for SAF.

Taxpayers can use the Provisional Emissions Rate process to obtain an emissions rate for fuel pathway and feedstock combinations not specified in the emissions rate table when guidance is published for the PER process. Guidance for the PER process is expected at a later date.

Outlining climate smart agriculture practices

The guidance released Friday states that the Treasury intends to propose rules for incorporating the emissions benefits from climate-smart agriculture (CSA) practices for cultivating domestic corn, soybeans, and sorghum as feedstocks for SAF and non-SAF transportation fuels. These options would be available to taxpayers after Treasury and the IRS propose regulations for the section 45Z credit, including rules for CSA, and the 45ZCF-GREET model is updated to enable calculation of the lifecycle greenhouse gas emissions rates for CSA crops, taking into account one or more CSA practices.    

CSA practices have multiple benefits, including lower overall GHG emissions associated with biofuels production and increased adoption of farming practices that are associated with other environmental benefits, such as improved water quality and soil health. Agencies across the Federal government have taken important steps to advance the adoption of CSA. In April, Treasury established a first-of-its-kind pilot program to encourage CSA practices within guidance on the section 40B SAF tax credit. Treasury has received and continues to consider substantial feedback from stakeholders on that pilot program. The U.S. Department of Agriculture invested more than $3 billion in 135 Partnerships for Climate-Smart Commodities projects. Combined with the historic investment of $19.5 billion in CSA from the Inflation Reduction Act, the department is estimated to support CSA implementation on over 225 million acres in the next 5 years as well as measurement, monitoring, reporting, and verification to better understand the climate impacts of these practices.

In addition, in June, the U.S. Department of Agriculture published a Request for Information requesting public input on procedures for reporting and verification of CSA practices and measurement of related emissions benefits, and received substantial input from a wide array of stakeholders. The USDA is currently developing voluntary technical guidelines for CSA reporting and verification. The Treasury and the IRS expect to consider those guidelines in proposing rules recognizing the benefits of CSA for purposes of the Section 45Z credit.

Continue Reading

Accounting

IRS and Treasury propose regs on 401(k) and 403(b) automatic enrollment, Roth IRA catchup contributions

Published

on

The Treasury Department and the Internal Revenue Service issued proposed regulations Friday for several provisions of the SECURE 2.0 Act, including ones related to automatic enrollment in 401(k) and 403(b) plans, and the Roth IRA catchup rule.

SECURE 2.0 Act passed at the end of 2022 and contained an extensive list of provisions related to retirement planning, like the original SECURE Act of 2019, with some being phased in over five years.

One set of proposed regulations involves provisions requiring newly-created 401(k) and 403(b) plans to automatically enroll eligible employees starting with the 2025 plan year. In general, unless an employee opts out, a plan needs to automatically enroll the employee at an initial contribution rate of at least 3% of the employee’s pay and automatically increase the initial contribution rate by one percentage point each year until it reaches at least 10% of pay. The requirement generally applies to 401(k) and 403(b) plans established after Dec. 29, 2022, the date the SECURE 2.0 Act became law, with exceptions for new and small businesses, church plans and governmental plans.

The proposed regulations include guidance to plan administrators for properly implementing this requirement and are proposed to apply to plan years that start more than six months after the date that final regulations are issued. Before the final regulations are applicable, plan administrators need to apply a reasonable, good faith interpretation of the statute.

Roth IRA catchup contributions

The Treasury and the IRS also issued proposed regulations Friday addressing several SECURE 2.0 Act provisions involving catch-up contributions, which are additional contributions under a 401(k) or similar workplace retirement plan that generally are allowed with respect to employees who are age 50 or older.

That includes proposed rules related to a provision requiring that catch-up contributions made by certain higher-income participants be designated as after-tax Roth contributions.

The proposed regulations provide guidance for plan administrators to implement and comply with the new Roth catch-up rule and reflect comments received in response to Notice 2023-62, issued in August 2023. 

The proposed regulations also provide guidance relating to the increased catch-up contribution limit under the SECURE 2.0 Act for certain retirement plan participants. Affected participants include employees between the ages of 60-63 and employees in newly established SIMPLE plans.

The IRS and the Treasury are asking for comments on both sets of proposed regulations. 

Continue Reading

Trending