Research map / Large language models / Retrieval and knowledge
Parameters vs. Context: Fine-Grained Control of Knowledge Reliance in Language Models
Baolong Bi, Shenghua Liu, Yiwei Wang, Yilong Xu, Junfeng Fang, Lingrui Mei, Xueqi Cheng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Confidence-Gain-based conflict detection before modulating tokens (ablation) | CK-PLUG · NQ counterfactual context (alpha=0.0, LLaMA2-7B-chat) | hit rate (%): 30.7 -> 74.8 | Improved |
| add conflict detection (ConD) (ablation) | CK-PLUG · NQ counterfactual context (alpha=1.0, Qwen-instruct) | hit rate (%): 51.3 -> 87.2 | Improved |
| add conflict detection (ConD) at moderate alpha (ablation) | CK-PLUG · NQ counterfactual context (alpha=0.5, Mistral-instruct) | hit rate (%): 91.3 -> 92.4 | No clear effect |
| set alpha from 0.0 to 1.0 to shift reliance toward parametric knowledge (ablation) | CK-PLUG · NQ counterfactual context (LLaMA3-8B-instruct) | MR (%): 9.9 -> 71.9 | Improved |
| alpha=0.5 as a neutral setting (ablation) | CK-PLUG · NQ counterfactual context (LLaMA3-8B-instruct) | MR (%): 43.5 -> 42.1 | No clear effect |
| use adaptive alpha from entropy ratio in general RAG | RAG with LLaMA3-8B-instruct · HotpotQA | accuracy: 43.1 -> 46.7 | Improved |
| use adaptive alpha from entropy ratio in general RAG | RAG with LLaMA2-7B-chat · FEVER | accuracy: 66.2 -> 72.6 | Improved |
| use adaptive alpha from entropy ratio in general RAG | RAG with Qwen2.5-7B-instruct · T-REX | accuracy: 61.6 -> 60.2 | Hurt |
| use adaptive alpha from entropy ratio in general RAG | RAG with LLaMA3-8B-instruct · FEVER | accuracy: 86.4 -> 86.1 | No clear effect |
| use adaptive alpha from entropy ratio in general RAG | RAG with LLaMA2-7B-chat · ELI5 (Rouge-L) | Rouge-L: 13.3 -> 14.1 | Improved |
Closest papers by meaning
- KnowPO: Knowledge-Aware Preference Optimization for Controllable Knowledge Selection in Retrieval-Augmented Language Models (AAAI 2025 · similarity 0.90)
- ParamMute: Suppressing Knowledge-Critical FFNs for Faithful Retrieval-Augmented Generation (NeurIPS 2025 · similarity 0.89)
- Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts (ICLR 2024 · similarity 0.89)
- Long-Context LLMs Meet RAG: Overcoming Challenges for Long Inputs in RAG (ICLR 2025 · similarity 0.88)
- Inference Scaling for Long-Context Retrieval Augmented Generation (ICLR 2025 · similarity 0.88)
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
