AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

ECD: Evidence-guided Contrastive Decoding in Retrieval-Augmented Generation with Accurate Knowledge Reference Adjustment

Yize Sui, Yan Xu, Kun Hu, Jing Ren, Wenjing Yang

AAAI 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
use dynamic evidence-based weight instead of fixed w_ECD = 0.6 (ablation)ECD (LLaMA-3-8B-Instruct) · PopQAEM: 42.73 -> 43.59Improved
use dynamic evidence-based weight instead of fixed w_ECD = 0.6 (ablation)ECD (LLaMA-3-8B-Instruct) · HotpotQAEM: 42.86 -> 47.23Improved
use dynamic weight instead of fixed w_ECD = 0.4 (ablation)ECD (LLaMA-3-8B-Instruct) · TriviaQAEM: 62.63 -> 67.89Improved
add retrieved context (w_ECD 0 to 1, RegCls to RegOpn) (ablation)ECD (LLaMA-3-8B-Instruct) · PopQAEM: 30.36 -> 40.03Improved
use evidence-based ECD instead of entropy-based ACD with noisy contextECD (LLaMA-3-8B-Instruct) · TriviaQA noisy contextEM: 56.38 -> 62.30Improved
use ECD instead of regular open-book decoding with noisy contextECD (LLaMA-3-8B-Instruct) · TriviaQA noisy contextEM: 37.55 -> 62.30Improved
use ECD instead of RegOpn with gold contextECD (LLaMA-3-8B-Instruct) · TriviaQA gold contextEM: 84.25 -> 85.36Improved
use ECD instead of second-best method on averageECD (LLaMA-3-8B-Instruct) · four-dataset averageEM: 49.58 -> 51.89Improved
use ECD instead of CDA on averageECD (LLaMA-3-8B-Instruct) · four-dataset averageEM: 47.33 -> 51.89Improved
use static context emphasis CAD instead of regular open-book decodingCAD (LLaMA-3-8B-Instruct) · four-dataset averageEM: 47.26 -> 44.66Hurt
increase top-K candidate tokens from 1 to 25 (ablation)ECD (LLaMA-2-7B-Chat) · TriviaQAEM: about 57 -> approximately 65Improved
increase top-K candidate tokens from 25 to 50 (ablation)ECD (LLaMA-3-8B-Instruct) · PopQAEM: roughly 43 -> about 44No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy