AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

``Someone Hid It!'': Query-Agnostic Black-Box Attacks on LLM-Based Retrieval

Jiate Li, Defu Cao, Li Li, Wei Yang, Yuehan Qin, Chenxiao Yu, Tiannuo Yang, Ryan A. Rossi, Yan Liu, Xiyang Hu, Yue Zhao

ICML 2026 · Adversarial robustness and security › Black-box and transfer attacks · approach: Data-free substitute training · 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
add word-embedding surrogate instead of last-hidden (ablation)DQ-A · BRIGHT Economics, Qwen1.5 victimRecall@25 after attack (lower is better): 0.289 -> 0.223Improved
add word-embedding surrogate instead of last-hidden (ablation)DQ-A · BRIGHT Robotics, jina victimRecall@25 after attack (lower is better): 0.245 -> 0.253Hurt
add adversarial query-document learning over plain sampled-query GCG (ablation)DQ-A-H vs S-GCG · BRIGHT Economics, Qwen1.5 victimRecall@25 after attack (lower is better): 0.299 -> 0.289Improved
add adversarial learning plus word-embedding over S-GCG (ablation)DQ-A · BRIGHT Robotics, Qwen1.5 victimRecall@25 after attack (lower is better): 0.181 -> 0.152Improved
increase sampled query count from 5 to 10 (ablation)DQ-A · BRIGHT, all retrieversattack-induced performance dropImproved
increase sampled query count from 10 to 15 (ablation)DQ-A · BRIGHT, all retrieversattack-induced performance dropMixed
increase injected token amount (ablation)DQ-A · BRIGHT, all retrieversattack-induced performance dropNo clear effect
inject tokens at the middle instead of beginning or end (ablation)DQ-A · BRIGHT, Embedding-Gemmaattack-induced performance dropHurt
add a reranker after the retrieverDQ-A · BRIGHT Economics, BGE-M3 + BGE-reranker-v2-M3NDCG@25: 0.139 -> 0.131No clear effect
reverse the attack losses to boost a documentDQ-A reversed · BRIGHT Economics, single picked documentRecall@50: 0.00227 -> 0.00601Improved
use Qwen3-Embedding-0.6B as victimDQ-A · BRIGHT averageRecall@25 after attack: 0.270 -> 0.264No 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