Search results for: fine-tuning

Authors: Haixu Yu, Wenhui Cao

Abstract:

Tailoring Large Language Models (LLMs) to perform legal reasoning has been a popular trend in the study of AI and law. Researchers have mainly employed two methods to unlock the potential of LLMs, namely by finetuning the LLMs to expand their knowledge of law and by restructuring the prompts (In-Context Learning) to optimize the LLMs’ understanding of the legal questions. Although claiming the finetuning and renovated prompting can make LLMs more competent in legal reasoning, most state-of-the-art studies show quite limited improvements of practicability. In this paper, drawing on the study of the complexity and low interpretability of legal texts, we propose a prompting strategy based on the Chain of Thought (CoT) method. Instead of merely instructing the LLM to reason “step by step”, the prompting strategy requires the tested LLM to identify the ambiguity in the questions as the first step and then allows the LLM to generate corresponding answers in line with different understandings of the identified terms as the following step. The proposed prompting strategy attempts to encourage LLMs to "interpret" the given text from various aspects. Experiments that require the LLMs to answer “case analysis” questions of bar examination with general LLMs such as GPT 4 and legal LLMs such as LawGPT show that the prompting strategy can improve LLMs’ ability to better understand complex legal texts.

Keywords: ambiguity-identification, prompt, large language model, legal text understanding

Procedia PDF Downloads 61

Authors: Aditya Karade, Sharada Falane, Dhananjay Deshmukh, Vijaykumar Mantri

Abstract:

Brain tumors pose a significant challenge in healthcare due to their complex nature and impact on patient outcomes. The application of deep learning (DL) algorithms in medical imaging have shown promise in accurate and efficient brain tumour detection. This paper explores the performance of various pre-trained DL models ResNet50, Xception, InceptionV3, EfficientNetB0, DenseNet121, NASNetMobile, VGG19, VGG16, and MobileNet on a brain tumour dataset sourced from Figshare. The dataset consists of MRI scans categorizing different types of brain tumours, including meningioma, pituitary, glioma, and no tumour. The study involves a comprehensive evaluation of these models’ accuracy and effectiveness in classifying brain tumour images. Data preprocessing, augmentation, and finetuning techniques are employed to optimize model performance. Among the evaluated deep learning models for brain tumour detection, ResNet50 emerges as the top performer with an accuracy of 98.86%. Following closely is Xception, exhibiting a strong accuracy of 97.33%. These models showcase robust capabilities in accurately classifying brain tumour images. On the other end of the spectrum, VGG16 trails with the lowest accuracy at 89.02%.

Keywords: brain tumour, MRI image, detecting and classifying tumour, pre-trained models, transfer learning, image segmentation, data augmentation

Procedia PDF Downloads 74

Authors: Jingwei Du

Abstract:

The need of optimizing air conditioning systems for existing buildings calls for control methods designed with energy-efficiency as a primary goal. The majority of current control methods boil down to two categories: empirical and model-based. To be effective, the former heavily relies on engineering expertise and the latter requires extensive historical data. Reinforcement Learning (RL), on the other hand, is a model-free approach that explores the environment to obtain an optimal control strategy often referred to as “policy”. This research adopts Proximal Policy Optimization (PPO) to improve chiller plant control, and enable the RL agent to collaborate with experienced engineers. It exploits the fact that while the industry lacks historical data, abundant operational data is available and allows the agent to learn and evolve safely under human supervision. Thanks to the development of language models, renewed interest in RL has led to modern, online, policy-based RL algorithms such as the PPO. This research took inspiration from “alignment”, a process that utilizes human feedback to finetune the pretrained model in case of unsafe content. The methodology can be summarized into three steps. First, an initial policy model is generated based on minimal prior knowledge. Next, the prepared PPO agent is deployed so feedback from both critic model and human experts can be collected for future finetuning. Finally, the agent learns and adapts itself to the specific chiller plant, updates the policy model and is ready for the next iteration. Besides the proposed approach, this study also used traditional RL methods to optimize the same simulated chiller plants for comparison, and it turns out that the proposed method is safe and effective at the same time and needs less to no historical data to start up.

Keywords: chiller plant, control methods, energy efficiency, proximal policy optimization, reinforcement learning

Procedia PDF Downloads 30

Authors: Dylan Elliott, Katya Pertsova

Abstract:

Preposition errors are some of the most common errors created by L2 speakers. In addition, improving error correction and detection methods remains an open issue in the realm of Natural Language Processing (NLP). This research investigates whether the bidirectional encoder representations from the transformers model (BERT) have the potential to correct preposition errors accurately enough to be useful in error correction software. This research finds that BERT performs strongly when the scope of its error correction is limited to preposition choice. The researchers used an open-source BERT model and over three hundred thousand edited sentences from Wikipedia, tagged for part of speech, where only a preposition edit had occurred. To test BERT’s ability to detect errors, a technique known as multi-level masking was used to generate suggestions based on sentence context for every prepositional environment in the test data. These suggestions were compared with the original errors in the data and their known corrections to evaluate BERT’s performance. The suggestions were further analyzed to determine if BERT more often agreed with the judgements of the Wikipedia editors. Both the untrained and fined-tuned models were compared. Finetuning led to a greater rate of error-detection which significantly improved recall, but lowered precision due to an increase in false positives or falsely flagged errors. However, in most cases, these false positives were not errors in preposition usage but merely cases where more than one preposition was possible. Furthermore, when BERT correctly identified an error, the model largely agreed with the Wikipedia editors, suggesting that BERT’s ability to detect misused prepositions is better than previously believed. To evaluate to what extent BERT’s false positives were grammatical suggestions, we plan to do a further crowd-sourcing study to test the grammaticality of BERT’s suggested sentence corrections against native speakers’ judgments.

Keywords: BERT, grammatical error correction, preposition error detection, prepositions

Procedia PDF Downloads 147