AI Paper Atlas

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

FORCE: Transferable Visual Jailbreaking Attacks via Feature Over-Reliance CorrEction

Runqi Lin, Alasdair Paren, Suqin Yuan, Muyang Li, Philip Torr, Adel Bibi, Tongliang Liu

CVPR 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: LLM jailbreak and injection defense

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 spectral rescaling alone to PGD (ablation)PGD visual jailbreak · MaliciousInstruct, Idefics3-8B-Llama3ASR (%): 53.00 -> 59.00Improved
add layer-aware feature regularisation alone to PGD (ablation)PGD visual jailbreak · MaliciousInstruct, Idefics3-8B-Llama3ASR (%): 53.00 -> 55.00Improved
combine layer regularisation with spectral rescaling (ablation)FORCE · MaliciousInstruct, Idefics3-8B-Llama3ASR (%): 53.00 -> 64.00Improved
apply FORCE instead of PGD against adapter-based MLLMsPGD visual jailbreak · MaliciousInstruct, Llava-v1.6-mistral-7bASR (%): 61.00 -> 69.00Improved
apply FORCE instead of PGD against early-fusion MLLMsPGD visual jailbreak · MaliciousInstruct, Qwen2.5-VL-7B-InstructASR (%): 5.00 -> 11.00Improved
apply FORCE instead of PGD against Gemini-2.5-ProPGD visual jailbreak · MaliciousInstructASR (%): 10.00 -> 10.00No clear effect
apply FORCE instead of PGD against Claude-Sonnet-4PGD visual jailbreak · AdvBenchASR (%): 1.00 -> 1.00No clear effect
apply FORCE in the zero-shot single-query settingPGD visual jailbreak · MaliciousInstruct, Llava-v1.6-mistral-7bASR (%): 26.00 -> 26.00No clear effect
use blank initial image with FORCE (ablation)FORCE · MaliciousInstruct, Idefics3-8B-Llama3ASR (%): 64.00 -> 83.00Improved
ensemble optimisation over two source modelsPGD visual jailbreak · MaliciousInstruct, Llava-v1.6ASR (%): 61.00 -> 59.00Hurt
add MI-FGSM momentum to PGDPGD visual jailbreak · MaliciousInstruct, Llava-v1.6ASR (%): 61.00 -> 49.00Hurt
add uniform or Gaussian noise defence to FORCE attacksFORCE · MaliciousInstruct, Idefics3-8B-Llama3ASR (%): 64.00 -> 62.00No clear effect

Later papers that compared with this method

1 later accepted paper reports FORCE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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