BigHugger
GH Repository · cvs-health

uqlm

[JMLR 2026] "UQLM: A Python Package for Uncertainty Quantification in Large Language Models"

stars
1,199
30-day movement
+21/day
Related entries
60
Connections
1
python/uvPythonai-evaluationai-safetyhallucination-detectionconfidence-scoreuncertainty-estimationhallucinationhallucination-evaluationotherhallucination-mitigationpythonllmllm-evaluationllm-hallucinationllm-safetyuncertainty-quantificationconfidence-estimation

UQLM is a Python package for uncertainty quantification in large language models, published in JMLR 2026. It provides confidence estimation and scoring tools aimed at detecting and evaluating LLM hallucinations.

Use it when you need to attach confidence scores to LLM outputs to flag likely hallucinations.

Use it to

  • Score LLM responses with confidence estimates
  • Detect likely hallucinations in model outputs
  • Evaluate uncertainty-quantification methods on LLM tasks
  • Integrate confidence scoring into LLM safety pipelines

For ML engineers and researchers evaluating LLM reliability

Role
other
Language
Python
Licence
Apache-2.0
Forks
132
Open issues
14
Last push
2026-09-14
Latest release
v0.1.0 · 2025-05-06
topicsllmuncertainty-quantificationhallucination-detectionconfidence-estimationai-evaluationai-safety