# THUDM/WebGLM

WebGLM: An Efficient Web-enhanced Question Answering System (KDD 2023)

Repository: https://github.com/THUDM/WebGLM
Canonical: https://ross.abutalabs.com/products/webglm
Language: Python
License: Apache-2.0
License Family: permissive
Topics: chatgpt, llm, rlhf, webglm
Last push: 2025-03-25T03:31:42+00:00

## Health v2 (maintenance only)
Score: 35/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 13, release rhythm 35, longevity 85
- inputs: {"age_days": 1193, "days_push": 526, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1602, forks 131 (observed 2026-08-28T04:05:09.747447+00:00)

## What it is
WebGLM is an efficient web-enhanced question answering system (KDD 2023) that combines a large language model with web search retrieval and human preference alignment (RLHF). It ships pretrained models (WebGLM-10B, WebGLM-2B), a QA dataset, and an official Python implementation with a demo.

## Use cases
- answer questions using up-to-date web search results with an llm
- build a retrieval-augmented question answering system
- run a web-enhanced chatbot grounded in search results
- fine-tune or evaluate llms with human preference data for qa
- use the webglm-qa dataset for research on web-enhanced qa
- deploy an open-source alternative to web-connected chatgpt

## When to choose
- you need an open-source, research-backed web-enhanced QA pipeline with pretrained checkpoints
- you want to reproduce or extend KDD 2023 research on retrieval-augmented QA with RLHF
- you need a smaller (2B) or larger (10B) model for grounded question answering on limited GPU hardware

## When to avoid
- you need a production-grade, actively maintained RAG framework rather than a research codebase
- you want a general-purpose chat assistant without web retrieval dependencies (e.g., SerpAPI key required)
- you need a lightweight plug-and-play library rather than a full system with model checkpoints

## Facets
- artifact type: application
- maturity: active
- function: rag, llm-inference, machine-learning, nlp, search-engine
- domain: large-language-models, artificial-intelligence
- platform: python
- tags: web-enhanced-qa, rlhf, question-answering, chatglm, research-project, kdd-2023, pretrained-models, retrieval-augmented-generation, natural-language-processing, gpu, docker, linux

## Member repositories
- THUDM/WebGLM (main) score 35

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:09.747447+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:52:36.666167+00:00, confidence not recorded.
  - readme: https://github.com/THUDM/WebGLM (fetched 2026-08-28T04:05:09.747447+00:00, sha 9e6aa8b28daa)
- Data as of 2026-08-30T08:39:29.467469+00:00.