TnT-LLM Implementation Details: Pipeline Design, Robustness, and Efficiency
22 Apr 2025
Explore the technical implementation of TnT-LLM, including pipeline design, techniques for ensuring robustness, and model selection strategies
TnT-LLM: Presenting Prompt Templates
22 Apr 2025
In this section, we present the prompt templates that were used for conversation summarization, label assignment, and taxonomy generation, updation, and review.
Additional Results: Cross-Lingual Taxonomy Evaluation and In-Depth Classification Analysis
22 Apr 2025
Explore further results on cross-lingual taxonomy performance, detailed annotation agreement analysis, and comprehensive classification metrics for TnT-LLM.
TnT-LLM Generated Taxonomies: User Intent and Conversation Domain Labels
22 Apr 2025
View the user intent and conversation domain taxonomies automatically generated by TnT-LLM and refined by human calibration for text classification.
TnT-LLM: Democratizing Text Mining with Automated Taxonomy and Scalable Classification
22 Apr 2025
TnT-LLM automates text mining, enabling efficient taxonomy generation, LLM-augmented classification, and democratized access to text insights
TnT-LLM: High-Quality Automated Text Mining and Efficient LLM-Augmented Classification
22 Apr 2025
Summary of TnT-LLM findings: automated taxonomy generation outperforms clustering, LLMs are effective evaluators (with caveats), and more
LLM-Augmented Text Classification: Distilling GPT-4 Labels into Efficient Classifiers
22 Apr 2025
We explore using GPT-4 for large-scale text annotation and train lightweight classifiers on these labels, comparing their accuracy and efficiency
TnT-LLM for Automated Taxonomy Generation: Outperforming Clustering Baselines
21 Apr 2025
We evaluate TnT-LLM for automated text taxonomy generation, comparing its accuracy and relevance against embedding-based clustering methods using human and LLM
TnT-LLM for User Intent and Conversational Domain Labeling in Bing Copilot
21 Apr 2025
We evaluate TnT-LLM on real-world Bing Copilot chat transcripts for user intent detection and conversational domain labeling and detailing our data sampling.