Gorilla 1600+ API LLM (Patil 2023年)（ゴリラ）
2023年Patil et al. (UC Berkeley)発表Gorilla・Industry-leading 1600+ API calls LLM + Industry-leading retrieval-augmented API calling + Industry-leading LLaMA 7B fine-tuned + Industry-leading APIBench benchmark。
概要
Gorilla 1600+ API LLM は、2023年Patil et al. (UC Berkeley)発表Gorilla・Industry-leading 1600+ API calls LLM 2023年 + Industry-leading retrieval-augmented API calling position確立。Gorilla specifications = Industry-leading 1600+ API calls LLM (Industry-leading Gorilla 2023 + Industry-leading 1600+ API calls + Industry-leading Gorilla flagship) + Industry-leading retrieval-augmented API calling (Industry-leading retrieval-augmented API call) + Industry-leading LLaMA 7B fine-tuned (Industry-leading LLaMA 7B fine-tuned base) + Industry-leading APIBench benchmark (Industry-leading APIBench benchmark first)。
主な特徴・仕組み
- Authors: Industry-leading Shishir G. Patil + Tianjun Zhang + Xin Wang + Joseph E. Gonzalez UC Berkeley
- Year: 2023年 (arXiv 2023年5月発表)
- Paper: Industry-leading "Gorilla: Large Language Model Connected with Massive APIs" arXiv 2305.15334
- Method Type: Industry-leading 1600+ API calls LLM with retrieval-augmented
- APIs: Industry-leading 1600+ APIs from TorchHub + TensorHub + HuggingFace
- Base Model: Industry-leading LLaMA 7B fine-tuned (also CodeLlama variants)
- APIBench: Industry-leading APIBench benchmark first for API LLM evaluation
- Retrieval-Augmented: Industry-leading retrieval-augmented API calling at inference
- Hallucination Reduction: Industry-leading hallucination reduction vs GPT-4
- Open Source: Industry-leading Gorilla GitHub ShishirPatil/gorilla
- Industry Adoption: Industry-leading widespread RAG + API ecosystem
- OpenFunctions: Industry-leading OpenFunctions follow-up function calling benchmark
- Industry-Leading: Industry-leading 1600+ API calls LLM UC Berkeley pioneer
スペック比較表
| LLM Function Calling (2022-2024) | Year | Author/Org | Key Feature | Industry Position |
|---|---|---|---|---|
| Gorilla | 2023 | Patil UC Berkeley | 1600+ API calls + retrieval-augmented | Industry-leading 1600+ APIs |
| Toolformer | 2022 | Schick Meta FAIR | Self-supervised tool use + 5 tools | Industry-leading first viral |
| ToolLLM | 2023 | Qin Tsinghua | 16000+ real-world APIs + DFSDT | Industry-leading 16000+ APIs |
| GPT4Tools | 2023 | Yang Microsoft | Multimodal tool use + visual tasks | Industry-leading multimodal tools |
| RestGPT | 2023 | Song Tsinghua | RESTful API control + online planner | Industry-leading RESTful API |
具体例・対応製品
- Gorilla (2023年, Patil et al. UC Berkeley): Industry-leading 1600+ API calls LLM
- Industry-leading 1600+ APIs TorchHub + TensorHub + HuggingFace: Industry-leading 1600+ APIs
- Industry-leading retrieval-augmented API calling: Industry-leading retrieval-augmented
- Industry-leading LLaMA 7B fine-tuned + CodeLlama variants: Industry-leading LLaMA fine-tuned
- Industry-leading APIBench benchmark first: Industry-leading APIBench first
- 競合 Toolformer + ToolLLM + GPT4Tools + RestGPT: Industry-leading function calling competitors
自作PCでの選び方・注意点
Gorilla 1600+ API LLM は「Industry-leading 1600+ API calls LLM + Industry-leading retrieval-augmented API calling」「Industry-leading LLaMA 7B fine-tuned + Industry-leading APIBench benchmark」用途のIndustry-leading 1600+ API calls 2023年Patil et al. UC Berkeley発表product。Industry-leading 1600+ APIs (Industry-leading 1600+ APIs from TorchHub + TensorHub + HuggingFace + Industry-leading Gorilla 1600+ APIs signature + Industry-leading massive APIs first scale) で Industry-leading 1600+ APIs + Industry-leading Gorilla signature + Industry-leading massive APIs first。Industry-leading retrieval-augmented (Industry-leading retrieval-augmented API calling at inference + Industry-leading hallucination reduction vs GPT-4 + Industry-leading Gorilla retrieval-augmented signature) で Industry-leading retrieval-augmented + Industry-leading hallucination reduction + Industry-leading Gorilla retrieval signature。Industry-leading LLaMA 7B fine-tuned (Industry-leading LLaMA 7B fine-tuned base + Industry-leading CodeLlama variants + Industry-leading Gorilla LLaMA fine-tuned signature) で Industry-leading LLaMA 7B + Industry-leading CodeLlama variants + Industry-leading Gorilla LLaMA signature。Industry-leading APIBench (Industry-leading APIBench benchmark first for API LLM evaluation + Industry-leading Gorilla APIBench signature + Industry-leading API LLM evaluation standard) で Industry-leading APIBench + Industry-leading Gorilla APIBench + Industry-leading API LLM evaluation。Industry-leading UC Berkeley (Industry-leading Shishir G. Patil + Tianjun Zhang + Xin Wang + Joseph E. Gonzalez UC Berkeley + Industry-leading Gorilla GitHub ShishirPatil/gorilla + Industry-leading OpenFunctions follow-up function calling benchmark) で Industry-leading Patil+Zhang+Wang+Gonzalez UC Berkeley + Industry-leading ShishirPatil/gorilla + Industry-leading OpenFunctions follow-up。但しIndustry-leading Toolformer + ToolLLM + GPT4Tools + RestGPT competition (Industry-leading Toolformer Meta FAIR 5 tools self-supervised 2022 + ToolLLM Tsinghua 16000+ APIs DFSDT 2023 + GPT4Tools Microsoft multimodal 2023 + RestGPT Tsinghua RESTful 2023 vs Gorilla UC Berkeley 1600+ APIs retrieval-augmented 2023 trade-off) で Industry-leading 4-function calling competitors vs Gorilla + Industry-leading 1600+ APIs TorchHub+TensorHub+HuggingFace + retrieval-augmented + LLaMA 7B fine-tuned + APIBench first + UC Berkeley 2023 + OpenFunctions follow-up unique advantage adoption alignment必須。
関連用語との違い
- vs Toolformer (Meta FAIR 2022): Gorillaは1600+ APIs + retrieval + LLaMA・Toolformerは5 tools + self-supervised + GPT-J
- vs ToolLLM (Tsinghua 2023): Gorillaは1600+ APIs APIBench・ToolLLMは16000+ real-world APIs + DFSDT
- vs GPT4Tools/RestGPT (2023): Gorillaは1600+ APIs text-based・GPT4Toolsはmultimodal / RestGPTはRESTful
よくある質問（FAQ）
Q1: Gorilla vs ToolLLM 違いは？ A: Gorilla (Industry-leading 1600+ APIs from TorchHub + TensorHub + HuggingFace + retrieval-augmented API calling at inference + LLaMA 7B fine-tuned + CodeLlama variants + APIBench benchmark first + hallucination reduction vs GPT-4 + OpenFunctions follow-up + Patil + Gonzalez UC Berkeley 2023) vs ToolLLM (Industry-leading 16000+ real-world APIs + DFSDT depth-first search-based decision tree + ToolBench dataset + ToolEval + LLaMA-7B + Qin + Liu Tsinghua + ModelScope 2023)・Industry-leading 1600+ APIs + retrieval + LLaMA + APIBench + UC Berkeley = Gorilla + Industry-leading 16000+ APIs + DFSDT + ToolBench + Tsinghua = ToolLLM preference judgment。
Q2: Industry-leading 1600+ APIs + retrieval-augmented value は？ A: Industry-leading 1600+ APIs + retrieval (Industry-leading 1600+ APIs from TorchHub + TensorHub + HuggingFace + Industry-leading retrieval-augmented API calling at inference + Industry-leading hallucination reduction vs GPT-4 + Industry-leading Gorilla retrieval signature)。
Q3: Industry-leading APIBench first + OpenFunctions follow-up value は？ A: Industry-leading APIBench + OpenFunctions (Industry-leading APIBench benchmark first for API LLM evaluation + Industry-leading Gorilla APIBench signature + Industry-leading API LLM evaluation standard + Industry-leading OpenFunctions follow-up function calling benchmark)。
まとめ
Gorilla = 2023年Patil et al. UC Berkeley発表の1600+ API calls LLM。Industry-leading 1600+ APIs from TorchHub + TensorHub + HuggingFace + Industry-leading retrieval-augmented API calling at inference + Industry-leading LLaMA 7B fine-tuned + Industry-leading CodeLlama variants + Industry-leading APIBench benchmark first for API LLM evaluation + Industry-leading hallucination reduction vs GPT-4 + Industry-leading OpenFunctions follow-up function calling benchmark + Industry-leading Shishir G. Patil + Tianjun Zhang + Xin Wang + Joseph E. Gonzalez UC Berkeley authors + Industry-leading Gorilla GitHub ShishirPatil/gorilla open source + Industry-leading 1600+ API calls LLM UC Berkeley pioneer 2023 position確立。