GH Repository · stanford-oval
WikiChat
WikiChat is an improved RAG. It stops the hallucination of large language models by retrieving data from a corpus.
- stars
- 1,618
- 30-day movement
- +10/day
- Related entries
- 60
- Connections
- 1
Pythonragemnlp2023factualityllmchatbotnlpnatural-language-processing
WikiChat is a retrieval-augmented generation (RAG) system that grounds large language model responses in data retrieved from a corpus, aiming to stop hallucination. It is a Python repository from stanford-oval, presented in connection with EMNLP 2023 and released under the Apache-2.0 license.
Reach for it when you need a chatbot whose answers are constrained by retrieved corpus data rather than the model's own generation.
Use it to
- Reduce hallucination in LLM chat responses
- Ground answers in a retrieved corpus
- Build a factuality-focused chatbot
- Study or extend RAG pipelines
- Reproduce EMNLP 2023 research
For NLP researchers and developers building factual LLM chatbots
- Role
- rag
- Language
- Python
- Licence
- Apache-2.0
- Forks
- 147
- Open issues
- 8
- Last push
- 2026-01-31
- Latest release
- v1.0 · 2024-08-23
topicsragfactualityllmchatbotnlphallucination