AI Paper Atlas

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

Tug-of-War No More: Harmonizing Accuracy and Robustness in Vision-Language Models via Stability-Aware Task Vector Merging

Junhao Dong, Xinghua Qu, Cong Zhang, Sua Qi Rong, Nguyen Duc Thai, Wenbo Pan, Xinfeng Li, Tongliang Liu, Piotr Koniusz, Yew-Soon Ong

ICLR 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Vision-language model robustness

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 PISTOLE merging instead of PMGPISTOLE · Zero-shot 14 datasets avg (eps 2/255)clean+robust sum: 110.20 -> 116.89Improved
add GISM stability masks to vanilla merging (ablation)PISTOLE · Zero-shot 14 datasets avgclean+robust sum: 111.11 -> 113.47Improved
add PCI re-weighting on top of GISM (ablation)PISTOLE · Zero-shot 14 datasets avgclean+robust sum: 113.47 -> 114.83Improved
add adversarial parameter trajectory (APT) on top of GISM and PCI (ablation)PISTOLE · Zero-shot 14 datasets avgclean+robust sum: 114.83 -> 116.89Improved
use naturally fine-tuned VLM instead of pre-trained zero-shot VLM as natural task vector (ablation)PISTOLE · Zero-shot 14 datasets avgclean+robust sum: 114.93 -> 116.89Improved
use pre-trained VLM as natural vector: robust accuracy (ablation)PISTOLE · Zero-shot 14 datasets avgrobust accuracy: 47.83 -> 47.65Hurt
use PISTOLE instead of PMG on ViT-H/14PISTOLE · Zero-shot 14 datasets, ViT-H/14clean+robust sum: 119.27 -> 126.06Improved
use PISTOLE instead of PMG at eps 4/255PISTOLE · Zero-shot avg, eps 4/255clean+robust sum: 92.69 -> 97.31Improved
use PISTOLE with LoRA instead of PMG with LoRAPISTOLE · Zero-shot avg, eps 3/255 (LoRA)clean+robust sum: 83.84 -> 88.82Improved
swap CLIP encoder for PISTOLE encoder in LLaVA 1.5 (COCO)PISTOLE · COCO captioning, LLaVA 1.5CIDEr sum: 156.4 -> 165.5Improved
use PISTOLE instead of FARE for hallucinationPISTOLE · POPE hallucinationF1: 80.8 -> 83.0Improved
use PISTOLE merging weight lambda 0.2 (ablation)PISTOLE · ImageNet adversariesloss-parameter curvatureImproved

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