Search for: All records

Finetuned large language models (LLMs) have shown remarkable performance in financial tasks, such as sentiment analysis and information retrieval. Due to privacy concerns, finetuning and deploying financial LLMs (FinLLMs) locally are crucial for institutions and individuals. In this paper, we employ quantized low-rank adaptation (QLoRA) to finetune FinLLMs, which leverage low-rank structure and quantization technique to significantly reduce computational requirements while maintaining model performance. We also employ data and pipeline parallelism to enable local finetuning on commodity GPUs. Experiments on financial datasets validate the efficacy of our approach in yielding notable improvements over the base models. 

Financial large language models (FinLLMs) have been applied to various tasks in business, finance, accounting, and auditing. Complex financial regulations and standards are critical to financial services, which LLMs must comply with. However, FinLLMs’ performance in understanding and interpreting financial regulations has rarely been studied. Therefore, we organize the Regulations Challenge, a shared task at COLING FinNLP-FNP-LLMFinLegal2025. It encourages the academic community to explore the strengths and limitations of popular LLMs. We create 9 novel tasks and corresponding question sets. In this paper, we provide an overview of these tasks and summarize participants’ approaches and results. We aim to raise awareness of FinLLMs’ professional capability in financial regulations. 

This study focuses on the effectiveness of learning transfer-focused or transfer-focused lab report writing instructional modules on engineering undergraduates’ lab report writing in entry-level engineering laboratory courses. The modules are novel due to their shared language to describe and reinforce foundational writing terms used by the writing faculty and are ready for immediate use by engineering lab instructors. Three different universities, consisting of a polytechnical university, a liberal arts-anchored private university, and a branch campus of a research-one land grant university, participated. Student lab report samples from six various sophomore-level engineering courses were collected. For the control group, none of the participating lab instructors accessed the transfer-focused modules (academic years of 2019-2020 and 2020-2021); sixty-four control group lab report samples were collected (n = 64). In the academic year 2021-2022, the lab instructors had access to the transfer-focused modules via the web to be encouraged to update their lab instructions; the experimental group lab report samples were collected from forty-two students (n = 42). Using defined writing outcomes, a panel of engineering lab instructors assessed the participating students’ early (one of the first reports in the class) and late lab reports (written near the end of the course). The lab report assessment analysis indicates that only 30% of the control group students could write their early lab reports at a satisfactory level, while 60% of the experimental group students reached a satisfactory level in their early labs. For both early and late lab reports, the experimental group students outperformed most outcomes over the control group. The notably improved outcomes were related to audience awareness, data presentation, data analysis, and data interpretation. The transfer-focused lab report writing pedagogy enhanced engineering undergraduates’ ability to engage in critical thinking practices, including analysis, interpretation, and evaluation of their lab data/products. Additionally, students appeared to improve their awareness of a technical audience, expecting engineering language, styles, and conventions commonly shared by writers in engineering.