自動化檢索系統Auto-RAG:提升LLM決策能力

Ai




Auto-RAG:一個以LLM決策能力為核心的自主迭代檢索模型

檢索增強生成(Retrieval Augmented Generation, RAG)是一種高效的解決方案,適用於知識密集型任務,能夠提高輸出的質量並使其更具確定性,並且減少幻覺現象。然而,RAG的輸出仍然可能雜亂,並且在面對複雜查詢時可能無法作出適當的回應。為了解決這一局限性,研究人員引入了迭代檢索更新,這種方法更新了重新檢索的結果,以滿足動態的信息需求。主要是為了解決知識稀疏性和在複雜查詢解決過程中的需求,重點在於兩個W——何時(When)和什麼(What)需要檢索。儘管有潛力,但現有的大多數方法仍然過於依賴以人為導向的規則和提示,這種依賴需要大量的人力,限制了大型語言模型(LLMs)的決策能力,實際上是對它們進行“餵養”,而非賦予自主性。

為了克服這些挑戰,中國科學院的研究團隊提出了Auto-RAG,一個自主的迭代檢索增強系統,重點在於LLM的決策能力。該系統包括LLM與檢索器之間的多輪對話。與傳統結果相比,Auto-RAG利用LLM的推理能力進行計劃、知識提取、查詢重寫,並不斷查詢檢索器,直到向用戶提供所需的解決方案。Auto-RAG引入了一個自動合成基於推理的指令的框架,使LLMs能夠在迭代RAG過程中獨立做出決策。這些指令允許在迭代RAG過程中以最小成本自動化LLM的決策。

作者將這一迭代過程概念化為LLM與檢索器之間的多輪互動,直到檢索器對信息的充分性有信心。每次迭代後,模型進行推理並調整檢索方法,以尋找適當的信息。該流程的核心部分無疑是推理部分。作者添加了三個不同的推理點,構成了一個檢索的思考鏈(Chain of Thought)。

檢索規劃:這是第一步,專注於與查詢相關的初步數據檢索。這一階段還包括評估模型是否需要更多的檢索,或者獲得的信息是否已經足夠。

信息提取:第二步使信息更具查詢特異性。在此步驟中,LLM從檢索的文檔中提取相關信息,以便最終答案的整理。這包括對關鍵信息的總結方法,以減少不準確性。

答案推理:該流程的最後一步包括LLM根據提取的信息形成最終決策。

此外,AutoRAG具有高度的動態性,因為它會根據查詢的複雜性自動調整迭代次數,省卻了計算的麻煩。這一框架的另一個優勢在於其用戶友好,並以自然語言編寫,提供了高程度的可解釋性。現在我們已經討論了Auto-RAG的功能及其對改善模型性能的重要性,接下來讓我們看看這一流程在實際測試中的表現。

研究團隊在監督環境下對LLMs進行微調,以實現自主檢索。他們為此從兩個數據集(Natural Questions和2WikiMultihopQA)中合成了10,000個基於推理的指令。該流程中使用的模型包括Llama-3-8B-Instruct(用於推理合成)和Qwen1.5-32B-Chat(用於重寫查詢)。數據在Llama模型上進行微調,以提高其無人干預的檢索效率。

為了測試所提方法的有效性,作者在六個代表性基準上對Auto-RAG框架進行了基準測試,這些基準涉及開放域和多跳回答數據集。多跳問答有多個子部分和多個查詢,使得應用標準RAG方法效率低下。結果驗證了Auto-RAG的主張,在數據受限的訓練中取得了良好的結果。選擇了一種無需管道的零-shot提示方法作為RAG的基準。作者還將Auto-RAG與一些多鏈接和基於思考鏈的方法進行比較,結果顯示Auto-RAG超越了其他模型。

結論:Auto-RAG在六個基準上通過自動化多步檢索過程任務並增強傳統RAG設置中的推理能力而達到了卓越的表現。它不僅提供了更好的結果,還在檢索過程中自我調整查詢,直到獲得所需的信息。

這項研究展示了如何通過技術的創新來提升大型語言模型的自主性和效率,這對未來的AI應用具有深遠的影響。Auto-RAG不僅是對現有技術的改進,更是推動了AI在知識檢索領域的發展。隨著這類技術的成熟,我們有理由相信,未來的AI系統將能更好地滿足用戶的需求,並提供更準確的答案。

以上文章由特價GPT API KEY所翻譯及撰寫。而圖片則由FLUX根據內容自動生成。

🎨 Nano Banana Pro 圖像生成器|打幾句說話就出圖

想畫人像、產品圖、插畫?SSFuture 圖像生成器支援 Flux Gemini Nano Banana Pro 改圖 / 合成, 打廣東話都得,仲可以沿用上一張圖繼續微調。

🆓 Flux 模型即玩,不用登入
🤖 登入後解鎖 Gemini 改圖
📷 支援上載參考圖再生成
⚡ 每天免費額度任你玩
✨ 即刻玩 AI 畫圖
{
  "image_generation_request": {
    "prompt": "Ultra-realistic portrait of a man walking toward the camera on an airport runway at night He wears a white long-sleeve shirt with sleeves rolled up and dress pants, shoes. The camera is very close, capturing his face sharply - textures of skin, smoke from his lips, and subtle reflections of firelight in his eyes. Behind him, slightly out of focus, a commercial airplane is burning intensely, with huge flames, roaring firestorms, and thick black smoke rising high. The fiery glow casts dramatic orange highlights on his shirt and face, creating deep shadows and a gritty, cinematic mood. Wet runway reflects the blaze, enhancing the dramatic atmosphere.",
    "dimensions": {
      "width": 1200,
      "height": 1200
    },
    "style_descriptors": [
      "Cinematic",
      "Photorealistic",
      "Gritty",
      "Dramatic Lighting",
      "Macro Photography",
      "8k resolution"
    ],
    "subject_details": {
      "action": "Walking toward camera, smoking",
      "clothing": "White long-sleeve shirt (rolled sleeves), dress pants, shoes",
      "facial_features": "Sharp focus, skin texture, firelight reflection in eyes"
    },
    "environment_details": {
      "location": "Airport runway at night",
      "background": "Commercial airplane burning, intense fire, thick black smoke, out of focus",
      "ground": "Wet runway, reflecting fire"
    }
  }
} Create a photorealistic movie-poster image for Zootopia 2. The scene shows a young Asian man and an oversized black shirt standing inside a crowded, dimly lit cinema. He’s flanked by Judy Hopps in full police uniform and Nick Wilde, both rendered in their official Disney CGI style. Capture the shot from a low angle so the characters feel larger than life. The cinema screen behind them is glowing with an intense action scene from the film. Let the light from the screen spill across their faces and clothing to create dramatic high contrast. The environment should feel alive with background silhouettes, soft haze, and the warm ambience of a packed theatre. Style the composition like an official Disney promotional poster: epic layout, sharp detail, high resolution, balanced framing, and clean edges. The man and the cinema should be fully photorealistic, while Judy and Nick retain their signature animated look.
At the bottom, include:
• The Disney logo
• The title ZOOTOPIA 2 in bold green and white
• The release line ONLY IN THEATERS NOVEMBER 26 Poster tone: cinematic, polished, high-end theatrical marketing. A young woman with fair skin is taking a selfie inside a fitness center (gym).

Main Subject: A young woman with long black hair in a messy bun. She is wearing a light cream or ivory long-sleeved sports zipper jacket and black tight sports shorts. She is also wearing long white socks with black stripes at the top and white sports shoes.

Expression and Pose: She is sitting on a weight training machine (apparently a leg press machine or similar) and holding a phone (with a leopard/leopard print case) to take a selfie in the mirror or using the front camera, with her face forward, a soft smile, and her face and eyes looking at the mirror.

Background (Gym): The background is dominated by modern gym equipment in dark gray and red. There are a few other unfocused people in the background, including a man on the left lifting weights or standing near a barbell, and another man in a green shirt standing near equipment. The gym floor appears dark.

Brief Prompt (Suitable for Image Search or Hashtags):

Selfie in the gym, young woman with messy bun and cream jacket, sitting on a training machine, aesthetic sporty, indoor lighting. Without changing her face.

發佈留言

發佈留言必須填寫的電子郵件地址不會公開。 必填欄位標示為 *

✨🎱 Instagram留言 →

AI即回覆下期六合彩預測

🧠 AI 根據統計數據即時生成分析

💬 只要留言,AI就會即刻覆你心水組合

🎁 完!全!免!費!快啲嚟玩!

IG 貼文示意圖 AI 即時回覆示意圖

下期頭獎號碼

📲 去 Instagram 即刻留言