type: "Research"
title: "Gorilla & APIBench (2023)"
description: "Fine-tuned LLM + document retriever; adapts to doc change, cuts hallucination."
resource: "https://arxiv.org/abs/2305.15334"
tags: [benchmark, tool-use, research]
generated: { by: human:crpage, at: 2026-07-09T09:44:00Z }
status: stable
sources: [{ id: primary, resource: "https://arxiv.org/abs/2305.15334" }]

Gorilla & APIBench (2023)

A genuine advance: pairing a fine-tuned model with a document retriever, claiming adaptation to test-time documentation change, and introducing APIBench. The key qualification: it improved retrieval-grounded syntax and reduced hallucinated usage but did not solve authorisation, stateful recovery, policy adherence or legal accountability.

Relates to: Discovery and retrieval, Failure taxonomy.