Polski

Posty z #functioncalling

#functioncalling

Publiczne posty w X, które używają #functioncalling.

Powiązane hashtagi

  2. SitePoint@sitepointdotcom
    Writing better prompts isn't enough anymore. The biggest AI breakthroughs now come from giving models access to tools, APIs, and real actions through function calling. That's what powers many of today's AI agents. 🔗 #AI #LLM #FunctionCalling #WebDev
    Pobierz
    001274
  6. Doug Finke@dfinke
    Via OpenAI: Fine-tune the completions API on function-calling for formatted responses without full function definitions. Save prompt tokens and streamline API interactions. Ideal for developers and data scientists. #OpenAI #FunctionCalling
    022466
  10. Evan Lin@LINE DevRel✓@Evan_Lin
    對於 #LLM #FunctionCalling 研究有熱情的人，可以跟我一起關注這個部分。 ---- 柏克萊功能調用排行榜（Berkeley Function-Calling Leaderboard，簡稱BFCL）最新評估結果出爐！這項排行榜首次全面評估了大型語言模型（LLMs）在應用程序和軟件中集成時調用功能和工具的能力。🚀 🥇排名第一的是OpenAI的GPT-4-0125-Preview（Prompt），整體準確率達到87%，每次調用成本為5.25美元，平均延遲1.97秒。 🥈緊隨其後的是Anthropic的Claude-3-Opus-20240229（Prompt），整體準確率為86.47%，每次調用成本為10.84美元，平均延遲4.61秒。 🥉第三名是OpenAI的GPT-4-1106-Preview（FC），整體準確率為84.35%，每次調用成本為5.07美元，平均延遲5.97秒。 BFCL的評估標準涵蓋了簡單函數、多重函數、平行函數和平行多重函數等類型，並支持REST API、SQL、Java和JavaScript等多種編程語言。這些評估不僅檢查了模型生成的API調用的正確性，還幫助識別了LLM在生成功能調用時所犯的一些常見錯誤，為了解當前模型的限制並提供改進的見解提供了重要資訊。 想了解更多關於大型語言模型在功能調用方面的性能表現？前往
    Pobierz
    0112852
  14. yamame@yamame61012288
    robo8080さんのAI_StackChan2をベースに万葉集読み等を加えたものに、motohさんのFunctionCallingで呼び出せるメモ機能を付け加え、やっと動いた(^^)/　フリーズすることがあったので、MAX_HISTORYやDynamicJsonDocument等の変数はroboさんの値に変更。 #AIｽﾀｯｸﾁｬﾝ2 #ChatGPT #FunctionCalling
    Pobierz
    15191.6K
  19. Salesforce AI Research✓@SFResearch
    Great explanation about #functioncalling, #xLAM, and their importance, by @salesforce CEO of AI, @clarashih
    Clara Shih @clarashih
    One powerful thing about LLMs is their ability to figure out where to look up data and take actions. This is known as #functioncalling— LLM converts natural language request into API call and/or database queries. Eg, sales rep might ask an AI copilot to cancel an order; the copilot will figure out that this lives inside an OMS application and then invokes the function to cancel. The #Trust stakes are higher for accuracy when it involves updating data and invoking actions, so I’m especially proud and excited to share that the new #xLAM-7b model from @Salesforce beats @OpenAI #GPT-4 in function calling. It comes in close second to @Anthropic’s #Claude 3.5-Sonnet while being *much* smaller and cheaper. Explore our hyper-efficient models and CRM LLM leaderboard: xLAM Models on @huggingface: Congrats @silviocinguetta @CaimingXiong & our brilliant @SFResearch Team! #LAM #LLM #AI #MachineLearning #genAI #agents #functioncalling #EinsteinTrustLayer
    0391.3K
  20. Clara Shih✓@clarashih
    One powerful thing about LLMs is their ability to figure out where to look up data and take actions. This is known as #functioncalling— LLM converts natural language request into API call and/or database queries. Eg, sales rep might ask an AI copilot to cancel an order; the copilot will figure out that this lives inside an OMS application and then invokes the function to cancel. The #Trust stakes are higher for accuracy when it involves updating data and invoking actions, so I’m especially proud and excited to share that the new #xLAM-7b model from @Salesforce beats @OpenAI #GPT-4 in function calling. It comes in close second to @Anthropic’s #Claude 3.5-Sonnet while being *much* smaller and cheaper. Explore our hyper-efficient models and CRM LLM leaderboard: xLAM Models on @huggingface: Congrats @silviocinguetta @CaimingXiong & our brilliant @SFResearch Team! #LAM #LLM #AI #MachineLearning #genAI #agents #functioncalling #EinsteinTrustLayer
    Pobierz
    5155916.3K
  21. ueponx@ueponx
    Function Callingの正体は「APIを2回叩く」だけだった🤗1回目でLLMに関数を選ばせて、実行して、2回目で結果をまとめてもらう。 この分業がわかると、LLMを組み込んだアプリの設計がイメージしやすくなります🙆‍♂️ #Ollama #FunctionCalling #ローカルLLM #AI #Python
    002197
  22. Pokaż więcej postów