UQLM

What is UQLM?

UQLM is a Code Generation tool. Uncertainty quantification for large language models. UQLM is free to use. It is aimed at researchers, data scientists and machine learning engineers.

UQLM is a Python package for uncertainty quantification in large language models. It provides tools for detecting hallucinations and evaluating model reliability through various scoring methods.

Built for researchers, data scientists and machine learning engineers.

Key capabilities

• Coding

Features

• Hallucination Detection

• Black-Box Scorers

  Uses consistency-based methods to evaluate model outputs

• White-Box Scorers

  Evaluates outputs using token probability analysis

• LLM-as-a-Judge Scorers

• Ensemble Scorers

• Long-Text Scorers

  Analyzes claims in extended text outputs

• Code Generation UQ

• Example Notebooks

  Includes tutorials and sample code for implementation

Use cases

• Detect unreliable outputs in language models
• Evaluate model reliability through consistency-based scoring
• Analyze claims in long text outputs for accuracy
• Quantify uncertainty in code generation tasks

Pros & cons

Pricing

Free.

Developer & API

FAQ