An official website of the United States government
This paper investigates the performance of a diverse set of large language models (LLMs) including leading closed-source (GPT-4, GPT-4o mini, Claude 3.5 Haiku) and open-source (Llama 3.1 70B, Llama 3.1 8B) models, alongside the earlier GPT-3.5 within the context of U.S. tax resolutions. AI-driven solutions like these have made substantial inroads into legal-critical systems with significant socio-economic implications. However, their accuracy and reliability have not been assessed in some legal domains, such as tax. Using the Volunteer Income Tax Assistance (VITA) certification tests—endorsed by the US Internal Revenue Service (IRS) for tax volunteering—this study compares these LLMs to evaluate their potential utility in assisting both tax volunteers as well as taxpayers, particularly those with low and moderate income. Since the answers to these questions are not publicly available, we first analyze 130 questions with the tax domain experts and develop the ground truths for each question. We then benchmarked these diverse LLMs against the ground truths using both the original VITA questions and syntactically perturbed versions (a total of 390 questions) to assess genuine understanding versus memorization/hallucinations. Our comparative analysis reveals distinct performance differences: closed-source models (GPT-4, Claude 3.5 Haiku, GPT-4o mini) generally demonstrated higher accuracy and robustness compared to GPT-3.5 and the open-source Llama models. For instance, on basic multiple-choice questions, top models like GPT-4 and Claude 3.5 Haiku achieved 83.33% accuracy, surpassing GPT-3.5 (54.17%) and the open-source Llama 3.1 8B (50.00%). These findings generally hold across both original and perturbed questions. However, the paper acknowledges that these developments are initial indicators, and further research is necessary to fully understand the implications of deploying LLMs in this domain. A critical limitation observed across all evaluated models was significant difficulty with open-ended questions, which require accurate numerical calculation and application of tax rules. We hope that this paper provides a means and a standard to evaluate the efficacy of current and future LLMs in the tax domain.
more »
« less
This content will become publicly available on December 1, 2026
Achieving GPT-4o level performance in astronomy with a specialized 8B-parameter large language model
Abstract AstroSage-Llama-3.1-8B is a domain-specialized natural-language AI assistant tailored for research in astronomy, astrophysics, cosmology, and astronomical instrumentation. Trained on the complete collection of astronomy-related arXiv papers from 2007 to 2024 along with millions of synthetically-generated question-answer pairs and other astronomical literature, AstroSage-Llama-3.1-8B demonstrates remarkable proficiency on a wide range of questions. AstroSage-Llama-3.1-8B scores 80.9% on the AstroMLab-1 benchmark, greatly outperforming all models—proprietary and open-weight—in the 8-billion parameter class, and performing on par with GPT-4o. This achievement demonstrates the potential of domain specialization in AI, suggesting that focused training can yield capabilities exceeding those of much larger, general-purpose models. AstroSage-Llama-3.1-8B is freely available, enabling widespread access to advanced AI capabilities for astronomical education and research.
more »
« less
- Award ID(s):
- 2406729
- PAR ID:
- 10612805
- Publisher / Repository:
- Scientific Reports
- Date Published:
- Journal Name:
- Scientific Reports
- Volume:
- 15
- Issue:
- 1
- ISSN:
- 2045-2322
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The authors introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments aimed at improving language models. DCLM provides a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants can experiment with dataset curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline, the authors find that model-based filtering is critical for assembling a high-quality training set. Their resulting dataset, DCLM-Baseline, enables training a 7B parameter model from scratch to achieve 64% 5-shot accuracy on MMLU with 2.6T training tokens. This represents a 6.6 percentage point improvement over MAP-Neo (the previous state-of-the-art in open-data LMs), while using 40% less compute. The baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% and 66%), and performs similarly on an average of 53 NLU tasks, while using 6.6x less compute than Llama 3 8B. These findings emphasize the importance of dataset design for training LMs and establish a foundation for further research on data curation.more » « less
-
Large Language Models (LLMs) are often augmented with external contexts, such as those used in retrieval-augmented generation (RAG). However, these contexts can be inaccurate or intentionally misleading, leading to conflicts with the model’s internal knowledge. We argue that robust LLMs should demonstrate situated faithfulness, dynamically calibrating their trust in external information based on their confidence in the internal knowledge and the external context to resolve knowledge conflicts. To benchmark this capability, we evaluate LLMs across several QA datasets, including a newly created dataset featuring in-the-wild incorrect contexts sourced from Reddit posts. We show that when provided with both correct and incorrect contexts, both open-source and proprietary models tend to overly rely on external information, regardless of its factual accuracy. To enhance situated faithfulness, we propose two approaches: Self-Guided Confidence Reasoning (SCR) and Rule-Based Confidence Reasoning (RCR). SCR enables models to self-access the confidence of external information relative to their own internal knowledge to produce the most accurate answer. RCR, in contrast, extracts explicit confidence signals from the LLM and determines the final answer using predefined rules. Our results show that for LLMs with strong reasoning capabilities, such as GPT-4o and GPT-4o mini, SCR outperforms RCR, achieving improvements of up to 24.2% over a direct input augmentation baseline. Conversely, for a smaller model like Llama-3-8B, RCR outperforms SCR. Fine-tuning SCR with our proposed Confidence Reasoning Direct Preference Optimization (CR-DPO) method improves performance on both seen and unseen datasets, yielding an average improvement of 8.9% on Llama-3-8B. In addition to quantitative results, we offer insights into the relative strengths of SCR and RCR. Our findings highlight promising avenues for improving situated faithfulness in LLMs.more » « less
-
Abstract The acquisition of complex astronomical data is accelerating, especially with newer telescopes producing ever more large-scale surveys. The increased quantity, complexity, and variety of astronomical data demand a parallel increase in skill and sophistication in developing, deciding, and deploying statistical methods. Understanding limitations and appreciating nuances in statistical and machine learning methods and the reasoning behind them is essential for improving data-analytic proficiency and acumen. Aiming to facilitate such improvement in astronomy, we delineate cautionary tales in statistics via six maxims, with examples drawn from the astronomical literature. Inspired by the significant quality improvement in business and manufacturing processes by the routine adoption of Six Sigma, we hope the routine reflection on these six maxims will improve the quality of both data analysis and scientific findings in astronomy.more » « less
-
This research was a qualitative transcendental phenomenological exploration of astronomy identity formation among astronomy majors and physics majors engaged in astronomy research. Participants (N=10), all of whom identified with traditionally marginalized groups in astronomy, were recruited from two large universities in New York State at different stages in their undergraduate careers. Social cognitive career theory and several STEM identity frameworks conceptually guided the analysis of astronomy identity for undergraduate majors and undergraduate astronomy researchers by exploring participants’ interest in, choice to study, and persistence in astronomy. Themes related to astronomy interest were popular culture and directly observing astronomical phenomena, while astronomy choice and persistence were facilitated by experiences in introductory coursework, recognition from faculty, and socializing with peers. The emergent astronomy identity framework was characterized by six distinct yet interrelated constructs: (1) interest in astronomy, typically rooted in observing naturally occurring phenomena and engaging with popular culture; (2) recognition from peers, experts, and families; (3) peer socialization about shared astronomy experiences; (4) competence, or the belief one can do the work of an astronomer; (5) sense of belonging within the astronomical community; and (6) astronomy career expectations. Findings from this research provide insights on factors that influence undergraduates in four-year colleges to study astronomy, and how students’ past experiences may lead to a natural interest in astronomy that may be fostered in secondary and post-secondary contexts. Implications suggest departments and institutions may facilitate the accessibility of astronomy at the collegiate level by promoting a more inclusive astronomy community, fostering interactions with astronomy faculty and graduate students, providing opportunities for undergraduate research, and communicating expectancy for astronomy-related future careers.more » « less
