Small finetuned LLMs for a diverse set of useful tasks

Code and dataset for our Bioinformatics 2022 paper: "A Benchmark for Automatic Medical Consultation System: Frameworks, Tasks and Datasets"

MEDIQA-Chat Shared Tasks @ ACL-ClinicalNLP 2023

EMNLP 2022: ClidSum: A Benchmark Dataset for Cross-Lingual Dialogue Summarization

Revisiting Cross-Lingual Summarization: A Corpus-based Study and A New Benchmark with Improved Annotation

[EMNLP 2023] FaMeSumm: Investigating and Improving Faithfulness of Medical Summarization. Support BART, PEGASUS, T5, mT5, BioBART, etc.

Dialogue Summarization application hosted using AWS and CICD deployment with docker and FASTAPI. Model card created in HuggingFace and a deployed on HuggingFace Spaces.

Fine Tuning pegasus and flan-t5 pre-trained language model on dialogsum datasets for conversation summarization to to optimize context window in RAG-LLMs

For this project, I fine-tuned two separate models for three tasks: document summarization, dialogue summarization and text classification

This notebook fine-tunes the FLAN-T5 model for dialogue summarization, comparing full fine-tuning with Parameter-Efficient Fine-Tuning (PEFT). It evaluates performance using ROUGE metrics, demonstrating PEFT's efficiency while achieving competitive results.

This repository explores enhancing dialogue summarization with commonsense knowledge through the SICK framework, evaluating models on dialogue datasets to assess commonsense's impact on summarization quality.

Fine-tuning FLAN-T5 with PPO and PEFT to generate less toxic text summaries. This notebook leverages Meta AI's hate speech reward model and utilizes RLHF techniques for improved safety.

Instruction fine tuning BART for Dialogue Summarization | IT4772E | NLP Project 20232

A dataset focused on summarization of dialogs, which represents the rich domain of Twitter customer care conversations.

GEN AI use case: dialogue summary. This notebook is extracted from the course Generative AI with Large Language Models. It is used to understand how input text can affect model performance.