From d9d612060934f12ebeb5b3d9557f03d9e556bb02 Mon Sep 17 00:00:00 2001 From: zouap Date: Wed, 12 Oct 2022 11:17:25 +0800 Subject: [PATCH] =?UTF-8?q?=E5=8E=BB=E6=8E=89ACTC=E6=8C=87=E6=A0=87?= =?UTF-8?q?=E5=80=BC=EF=BC=8C=E8=AF=84=E6=B5=8B=E7=BB=93=E6=9E=9C=E4=BC=9A?= =?UTF-8?q?=E6=8A=A5=E4=BB=A3=E7=A0=81=E5=BC=82=E5=B8=B8=E3=80=82?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: zouap --- templates/repo/modelsafety/indicators_descr.tmpl | 7 ++++--- templates/repo/modelsafety/new.tmpl | 4 ++-- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/templates/repo/modelsafety/indicators_descr.tmpl b/templates/repo/modelsafety/indicators_descr.tmpl index ac9b934a8..35c166e2e 100644 --- a/templates/repo/modelsafety/indicators_descr.tmpl +++ b/templates/repo/modelsafety/indicators_descr.tmpl @@ -25,14 +25,15 @@ ALDp Average LpDistortion：为所有攻击成功的对抗样本的平均归一化Lp失真度，ALDp值越小，对抗样本的不可感知性越强。 + ACAC Average Confidence of Adversarial Class(ACAC)：这个数值越高，攻击算法的攻击能力越强。 - - + + PSD Perturbation Sensitivity Distance：用于评测人类对扰动的感知能力。该指标值越大，说明评测结果越差。 diff --git a/templates/repo/modelsafety/new.tmpl b/templates/repo/modelsafety/new.tmpl index 4472a310c..cc3e42ab2 100644 --- a/templates/repo/modelsafety/new.tmpl +++ b/templates/repo/modelsafety/new.tmpl @@ -255,8 +255,8 @@ - - + +