Agents & Tool Use · 2021
WebGPT: Browser-assisted Question-answering with Human Feedback
Fine-tuned a language model to answer questions by browsing the web and citing sources, using human feedback to reward well-supported answers and reducing unsupported claims.
Editorial record
Plain-language summary
WebGPT gives GPT-3 a text-based browser and trains it with human comparisons to search, navigate, and quote evidence, then optimizes against a learned reward model. The result answers long-form questions with references that raters prefer, at the cost of sometimes over-trusting sources. It was an early demonstration of tool use and retrieval-augmented answering with RLHF.
Knowledge graph
Relationships
Antecedents
Conceptual ancestorEvidence: Strongly supported
ReAct: Synergizing Reasoning and Acting in LMs
Tool-augmented answering preceded the ReAct agent pattern
book
Descendants
Depends onEvidence: Direct
Language Models are Few-Shot Learners (GPT-3)
WebGPT fine-tunes GPT-3 to browse and answer with citations
P-520
Source record
Provenance
- Record ID
- P-520
- Record created
- 2026-07-13
- Last reviewed
- 2026-07-14
- Record version
- 2
- https://arxiv.org/abs/2112.09332
- arXiv:2112.09332
Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.