AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Exposing the Cracks: Vulnerabilities of Retrieval-Augmented LLM-based Machine Translation

Yanming Sun, Runzhe Zhan, Chi Seng Cheang, Han Wu, Xuebo Liu, Yuyao Niu, Fengying Ye, Kaixin Lan, Lidia S. Chao, Derek F. Wong

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Long-context and personalized retrieval · 2 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
replace gold meaning with opposite-meaning retrieved contextQwen2.5-7B-Instruct · Idiom translation avg 10 pairsFidelity: 2.3 -> 0.8Hurt
replace gold meaning with structure-perturbed gold meaningQwen2.5-7B-Instruct · Idiom translation avg 10 pairsFidelity: 2.3 -> 2.1No clear effect
add literal-translation noise instead of no contextQwen2.5-7B-Instruct · Idiom translation avg 10 pairsFidelity: 1.2 -> 1.4No clear effect
use thinking mode instead of non-thinkingQwen3-8B · Idiom translation, opposite-meaning noiseFidelity: 0.8 -> 0.4Hurt
use thinking mode instead of non-thinkingQwen3-8B · Idiom translation, semantic-perturbed noiseFidelity: 1.1 -> 0.9Hurt
add ALI fine-tuning (opposite-meaning noise training)Qwen2.5-7B-Instruct · Fr→En, opposite noiseFidelity: 0.7 -> 1.8Improved
add ALI fine-tuningQwen2.5-7B-Instruct · Hi→En, opposite noiseFidelity: 0.3 -> 0.8Improved
add ALI fine-tuningQwen2.5-7B-Instruct · Hi→En, gold meaningFidelity: 2.1 -> 1.7Hurt
add CK-PLUG training-free decodingQwen2.5-7B-Instruct · Fr→En, opposite noiseFidelity: 0.7 -> 1.3Improved
add CK-PLUG training-free decodingQwen2.5-7B-Instruct · Hi→En, gold meaningFidelity: 2.1 -> 1.5Hurt
use CDA instead of ALI fine-tuning (ablation)Qwen2.5-7B-Instruct · Hi→En, opposite noiseFidelity: 0.7 vs 0.8No clear effect
use ALI fine-tuning, evaluated on syntactic noiseQwen2.5-7B-Instruct · Fr→En, NstructFidelity: 2.3 -> 2.2No 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