Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2509.06472 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909778137055232 |
|---|---|
| author | Jin, Haoxiang Li, Ronghan Lu, Zixiang Miao, Qiguang |
| author_facet | Jin, Haoxiang Li, Ronghan Lu, Zixiang Miao, Qiguang |
| contents | Large Language Models (LLMs) often generate inaccurate responses (hallucinations) when faced with questions beyond their knowledge scope. Retrieval-Augmented Generation (RAG) addresses this by leveraging external knowledge, but a critical challenge remains: determining whether retrieved contexts effectively enhance the model`s ability to answer specific queries. This challenge underscores the importance of knowledge boundary awareness, which current methods-relying on discrete labels or limited signals-fail to address adequately, as they overlook the rich information in LLMs` continuous internal hidden states. To tackle this, we propose a novel post-retrieval knowledge filtering approach. First, we construct a confidence detection model based on LLMs` internal hidden states to quantify how retrieved contexts enhance the model`s confidence. Using this model, we build a preference dataset (NQ_Rerank) to fine-tune a reranker, enabling it to prioritize contexts preferred by the downstream LLM during reranking. Additionally, we introduce Confidence-Based Dynamic Retrieval (CBDR), which adaptively triggers retrieval based on the LLM`s initial confidence in the original question, reducing knowledge conflicts and improving efficiency. Experimental results demonstrate significant improvements in accuracy for context screening and end-to-end RAG performance, along with a notable reduction in retrieval costs while maintaining competitive accuracy. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_06472 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Rethinking LLM Parametric Knowledge as Post-retrieval Confidence for Dynamic Retrieval and Reranking Jin, Haoxiang Li, Ronghan Lu, Zixiang Miao, Qiguang Information Retrieval Large Language Models (LLMs) often generate inaccurate responses (hallucinations) when faced with questions beyond their knowledge scope. Retrieval-Augmented Generation (RAG) addresses this by leveraging external knowledge, but a critical challenge remains: determining whether retrieved contexts effectively enhance the model`s ability to answer specific queries. This challenge underscores the importance of knowledge boundary awareness, which current methods-relying on discrete labels or limited signals-fail to address adequately, as they overlook the rich information in LLMs` continuous internal hidden states. To tackle this, we propose a novel post-retrieval knowledge filtering approach. First, we construct a confidence detection model based on LLMs` internal hidden states to quantify how retrieved contexts enhance the model`s confidence. Using this model, we build a preference dataset (NQ_Rerank) to fine-tune a reranker, enabling it to prioritize contexts preferred by the downstream LLM during reranking. Additionally, we introduce Confidence-Based Dynamic Retrieval (CBDR), which adaptively triggers retrieval based on the LLM`s initial confidence in the original question, reducing knowledge conflicts and improving efficiency. Experimental results demonstrate significant improvements in accuracy for context screening and end-to-end RAG performance, along with a notable reduction in retrieval costs while maintaining competitive accuracy. |
| title | Rethinking LLM Parametric Knowledge as Post-retrieval Confidence for Dynamic Retrieval and Reranking |
| topic | Information Retrieval |
| url | https://arxiv.org/abs/2509.06472 |