AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Rethinking Deep Alignment Through the Lens of Incomplete Safety Learning

Thong Bach, Dung Nguyen, Thao Minh Le, Truyen Tran

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Adversarial prompts for language models · 1 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 targeted completion (adaptive base-favored token penalty + hybrid teacher distillation)Llama-2-7B-Chat · GCG attackattack success rate (%): 51.0 -> 0.4Improved
add targeted completionLlama-3.1-8B-Instruct · Prefill attack (AdvBench)attack success rate (%): 90.1 -> 14.8Improved
add targeted completionQwen-2.5-7B-Instruct · Prefill attack (AdvBench)attack success rate (%): 85.9 -> 44.3Improved
add targeted completionfour model families · MMLU, ARC-C, BoolQ, HellaSwag, Winograndeaccuracy: MMLU 68.0 -> 68.9 (Llama-3.1-8B)No clear effect
add targeted completion after Dolly fine-tuningLlama-3.1-8B LoRA · Dolly fine-tuning recoveryHRR (%): 25.5 -> 1.0Improved
use targeted completion instead of Vaccine, SaLoRA, Safe LoRALoRA fine-tuned models · Prefill attack after fine-tuningattack success rate (%): 69.3-73.4 -> 44.5Improved
add inference-time contrastive decoding penalizing base-favored tokensLlama-3.1-8B-Instruct · Prefill attack (4-token)attack success rate (%): 47.5 -> 0.2Improved
add targeted completionQwen-3-8B-Instruct · Prefill attack, 384 AdvBench promptssafe responses with reasoning (count): 9 -> 196Improved
add targeted completion versus safety augmentationLlama-2-7B-Chat · GCG attackattack success rate (%): 1.6 -> 0.4No clear effect
add targeted completion after Dolly fine-tuningQwen-2.5-7B LoRA · GSM8Kaccuracy: Llama-3.1-8B 72.4 -> 78.1; Qwen-2.5-7B 60.2 -> 73.1 vs Base 76.4Mixed

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