LIT @ UMich's repositories
vlog_action_recognition
Identifying Visible Actions in Lifestyle Vlogs
In-the-wild-QA
In-the-wild Question Answering
micromodels
Micromodels -- A framework for accurate, explainable, data efficient, and reusable NLP models.
Reddit-Detect-Inspiration
Inspirational post ids collected from Reddit using pushift.io and RoBERTa
michigannlp.github.io
LIT group website
visual_diversity_budget
Annotations on a Budget: Leveraging Geo-Data Similarity to Balance Model Performance and Annotation Cost
video-fill-in-the-blank
FIBER data and code.
Scalable-VLM-Probing
Probe Vision-Language Models
textgraphs
Website of TextGraphs: Graph-based Algorithms for Natural Language Processing
Bridging_the_Digital_Divide
Code for analyses included in Digital Divide paper
depression_synthetic_data
Can LMs generate useful synthetic data for the mental health domain?
vlog_action_co-occurrence
Human action co-occurrence identification in lifestyle vlogs
vlog_action_localization
Localizing narrated human activities in lifestyle vlogs.
cross_inspiration
Cross-cultural Inspiration Detection and Analysis in Real and LLM-generated Social Media Data
girls-encoded-website
Girls Encoded website
multilingual_reviews_deception
Multilingual Deception Detection of GPT-generated Hotel Reviews
task-adaptive_tokenization
A code implementation for the EMNLP 2023 paper "Enhancing Long-form Text Generation Efficacy with Task-adaptive Tokenization"
AnnDisagree
Annotation Embeddings website code