Update app.py

app.py

@@ -1,7 +1,7 @@
 # ================================================================
-# 教育大模型MIA攻防研究 - Gradio演示系统 v6.2 学术巅峰版 (苹果风)
-# 整合了双雷达图 + 算法流程图 + 伪代码 + 详尽数据分析 + 完整结论
-# ！！！全局修正：所有 e 替换为 ε / $\epsilon$，所有 s 替换为 σ / $\sigma$ ！！！
+# 教育大模型MIA攻防研究 - Gradio演示系统 v8.0 无死角终极版
+# 1. 修复：彻底弃用容易渲染失败的 LaTeX，全面采用原生 Unicode ε 和 σ
+# 2. 修复：补齐“输出扰动”在详细分析表格中丢失的 TPR@1%FPR 和 Loss差距
 # ================================================================
 
 import os
@@ -59,12 +59,11 @@ except FileNotFoundError:
     for s in [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]:
         k = f"perturbation_{s}"
         perturb_results[k] = {m: v*0.85 for m, v in mia_results["baseline"].items()} 
-        # 模拟方差变大
         perturb_results[k]["member_loss_std"] = np.sqrt(0.03**2 + s**2)
         perturb_results[k]["non_member_loss_std"] = np.sqrt(0.03**2 + s**2)
 
 # ================================================================
-# 全局图表配置 - 简约苹果风
+# 全局图表配置
 # ================================================================
 COLORS = {
     'bg': '#FFFFFF',           
@@ -81,8 +80,6 @@ COLORS = {
     'ls_colors': ['#A0C4FF', '#70A1FF', '#478EFF', '#007AFF'], 
     'op_colors': ['#98F5E1', '#6EE7B7', '#34D399', '#10B981', '#059669', '#047857'], 
 }
-
-# 图表宽度配置 (为了适配双雷达图)
 CHART_W = 14
 
 def apply_light_style(fig, ax_or_axes):
@@ -102,15 +99,15 @@ def apply_light_style(fig, ax_or_axes):
         ax.set_axisbelow(True) 
 
 # ================================================================
-# 提取指标的辅助函数 (核心替换：使用 LaTeX \epsilon 和 \sigma 画图)
+# 核心修复 1：使用标准 Unicode ε 和 σ，告别乱码
 # ================================================================
 LS_KEYS = ["baseline", "smooth_eps_0.02", "smooth_eps_0.05", "smooth_eps_0.1", "smooth_eps_0.2"]
-LS_LABELS_PLOT = ["Baseline", r"LS($\epsilon$=0.02)", r"LS($\epsilon$=0.05)", r"LS($\epsilon$=0.1)", r"LS($\epsilon$=0.2)"]
+LS_LABELS_PLOT = ["Baseline", "LS(ε=0.02)", "LS(ε=0.05)", "LS(ε=0.1)", "LS(ε=0.2)"]
 LS_LABELS_MD = ["基线(Baseline)", "LS(ε=0.02)", "LS(ε=0.05)", "LS(ε=0.1)", "LS(ε=0.2)"]
 
 OP_SIGMAS = [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]
 OP_KEYS = [f"perturbation_{s}" for s in OP_SIGMAS]
-OP_LABELS_PLOT = [r"OP($\sigma$={})".format(s) for s in OP_SIGMAS]
+OP_LABELS_PLOT = [f"OP(σ={s})" for s in OP_SIGMAS]
 OP_LABELS_MD = [f"OP(σ={s})" for s in OP_SIGMAS]
 
 ALL_KEYS = LS_KEYS + OP_KEYS
@@ -130,12 +127,8 @@ bl_acc = gu("baseline")
 bl_m_mean = gm("baseline", "member_loss_mean")
 bl_nm_mean = gm("baseline", "non_member_loss_mean")
 
-TYPE_CN = {'calculation': '基础计算', 'word_problem': '应用题',
-           'concept': '概念问答', 'error_correction': '错题订正'}
+TYPE_CN = {'calculation': '基础计算', 'word_problem': '应用题', 'concept': '概念问答', 'error_correction': '错题订正'}
 
-# ================================================================
-# 效用评估题库 
-# ================================================================
 np.random.seed(777)
 EVAL_POOL = []
 _types = ['calculation']*120 + ['word_problem']*90 + ['concept']*60 + ['error_correction']*30
@@ -149,8 +142,7 @@ for _i in range(300):
         else: _q,_ans=f"{_a} x {_b} = ?",str(_a*_b)
     elif _t == 'word_problem':
         _a,_b = int(np.random.randint(5,200)), int(np.random.randint(3,50))
-        _tpls = [(f"{_a} apples, ate {_b}, left?",str(_a-_b)),
-                 (f"{_a} per group, {_b} groups, total?",str(_a*_b))]
+        _tpls = [(f"{_a} apples, ate {_b}, left?",str(_a-_b)), (f"{_a} per group, {_b} groups, total?",str(_a*_b))]
         _q,_ans = _tpls[_i%len(_tpls)]
     elif _t == 'concept':
         _cs = [("area","Area = space occupied by a shape"),("perimeter","Perimeter = total boundary length")]
@@ -165,41 +157,27 @@ for _i in range(300):
     EVAL_POOL.append(item)
 
 # ================================================================
-# 图表绘制函数 (全面应用 LaTeX 标签渲染)
+# 图表绘制函数 
 # ================================================================
 def fig_gauge(loss_val, m_mean, nm_mean, thr, m_std, nm_std):
-    fig, ax = plt.subplots(figsize=(10, 2.6))
-    fig.patch.set_facecolor(COLORS['bg'])
-    ax.set_facecolor(COLORS['panel'])
-    
-    xlo = min(m_mean - 3.0 * m_std, loss_val - 0.005)
-    xhi = max(nm_mean + 3.0 * nm_std, loss_val + 0.005)
-    
-    ax.axvspan(xlo, thr, alpha=0.2, color=COLORS['accent'])
-    ax.axvspan(thr, xhi, alpha=0.2, color=COLORS['danger'])
-    
+    fig, ax = plt.subplots(figsize=(10, 2.6)); fig.patch.set_facecolor(COLORS['bg']); ax.set_facecolor(COLORS['panel'])
+    xlo = min(m_mean - 3.0 * m_std, loss_val - 0.005); xhi = max(nm_mean + 3.0 * nm_std, loss_val + 0.005)
+    ax.axvspan(xlo, thr, alpha=0.2, color=COLORS['accent']); ax.axvspan(thr, xhi, alpha=0.2, color=COLORS['danger'])
     ax.axvline(m_mean, color=COLORS['accent'], lw=2, ls=':', alpha=0.8, zorder=2)
     ax.text(m_mean - 0.002, 1.02, f'Member Mean\n{m_mean:.4f}', ha='right', va='bottom', fontsize=9, color=COLORS['accent'], transform=ax.get_xaxis_transform())
-
     ax.axvline(nm_mean, color=COLORS['danger'], lw=2, ls=':', alpha=0.8, zorder=2)
     ax.text(nm_mean + 0.002, 1.02, f'Non-Member Mean\n{nm_mean:.4f}', ha='left', va='bottom', fontsize=9, color=COLORS['danger'], transform=ax.get_xaxis_transform())
-
     ax.axvline(thr, color=COLORS['text_dim'], lw=2.5, ls='--', zorder=3)
     ax.text(thr, 1.25, f'Threshold\n{thr:.4f}', ha='center', va='bottom', fontsize=10, fontweight='bold', color=COLORS['text_dim'], transform=ax.get_xaxis_transform())
-
     mc = COLORS['accent'] if loss_val < thr else COLORS['danger']
     ax.plot(loss_val, 0.5, marker='o', ms=16, color='white', mec=mc, mew=3, zorder=5, transform=ax.get_xaxis_transform())
     ax.text(loss_val, 0.75, f'Current Loss\n{loss_val:.4f}', ha='center', fontsize=11, fontweight='bold', color=mc, transform=ax.get_xaxis_transform())
-
     ax.text((xlo+thr)/2, 0.25, 'MEMBER', ha='center', fontsize=12, color=COLORS['accent'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
     ax.text((thr+xhi)/2, 0.25, 'NON-MEMBER', ha='center', fontsize=12, color=COLORS['danger'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
-
     ax.set_xlim(xlo, xhi); ax.set_yticks([])
     for s in ax.spines.values(): s.set_visible(False)
-    ax.spines['bottom'].set_visible(True); ax.spines['bottom'].set_color(COLORS['grid'])
-    ax.tick_params(colors=COLORS['text_dim'], width=1)
-    ax.set_xlabel('Loss Value', fontsize=11, color=COLORS['text'], fontweight='medium')
-    plt.tight_layout(pad=0.5)
+    ax.spines['bottom'].set_visible(True); ax.spines['bottom'].set_color(COLORS['grid']); ax.tick_params(colors=COLORS['text_dim'], width=1)
+    ax.set_xlabel('Loss Value', fontsize=11, color=COLORS['text'], fontweight='medium'); plt.tight_layout(pad=0.5)
     return fig
 
 def fig_auc_bar():
@@ -221,66 +199,25 @@ def fig_auc_bar():
 
 def fig_radar():
     ms = ['AUC', 'Atk Acc', 'Prec', 'Recall', 'F1', 'TPR@5%', 'TPR@1%', 'Gap']
-    mk = ['auc', 'attack_accuracy', 'precision', 'recall', 'f1',
-          'tpr_at_5fpr', 'tpr_at_1fpr', 'loss_gap']
-    N = len(ms)
-    ag = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist() + [0]
-
-    fig, axes = plt.subplots(1, 2, figsize=(CHART_W + 2, 7),
-                              subplot_kw=dict(polar=True))
-    fig.patch.set_facecolor('white')
-
-    # --- 左图: 5个标签平滑模型 (替换LaTeX) ---
-    ls_cfgs = [
-        ("Baseline",             "baseline",         '#F04438'),
-        (r"LS($\epsilon$=0.02)", "smooth_eps_0.02",  '#B2DDFF'),
-        (r"LS($\epsilon$=0.05)", "smooth_eps_0.05",  '#84CAFF'),
-        (r"LS($\epsilon$=0.1)",  "smooth_eps_0.1",   '#2E90FA'),
-        (r"LS($\epsilon$=0.2)",  "smooth_eps_0.2",   '#7A5AF8'),
-    ]
-
-    # --- 右图: Baseline + 6个输出扰动 (替换LaTeX) ---
-    op_cfgs = [
-        ("Baseline",               "baseline",            '#F04438'),
-        (r"OP($\sigma$=0.005)",    "perturbation_0.005",  '#A6F4C5'),
-        (r"OP($\sigma$=0.01)",     "perturbation_0.01",   '#6CE9A6'),
-        (r"OP($\sigma$=0.015)",    "perturbation_0.015",  '#32D583'),
-        (r"OP($\sigma$=0.02)",     "perturbation_0.02",   '#12B76A'),
-        (r"OP($\sigma$=0.025)",    "perturbation_0.025",  '#039855'),
-        (r"OP($\sigma$=0.03)",     "perturbation_0.03",   '#027A48'),
-    ]
-
-    for ax_idx, (ax, cfgs, title) in enumerate([
-        (axes[0], ls_cfgs, 'Label Smoothing (5 models)'),
-        (axes[1], op_cfgs, 'Output Perturbation (7 configs)')
-    ]):
-        ax.set_facecolor('white')
+    mk = ['auc', 'attack_accuracy', 'precision', 'recall', 'f1', 'tpr_at_5fpr', 'tpr_at_1fpr', 'loss_gap']
+    N = len(ms); ag = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist() + [0]
+    fig, axes = plt.subplots(1, 2, figsize=(CHART_W + 2, 7), subplot_kw=dict(polar=True)); fig.patch.set_facecolor('white')
 
-        mx = []
-        for i, m_key in enumerate(mk):
-            val_max = max(gm(k, m_key) for _, k, _ in cfgs)
-            mx.append(val_max if val_max > 0 else 1)
+    # 完全使用 Unicode 希腊字母
+    ls_cfgs = [("Baseline", "baseline", '#F04438'), ("LS(ε=0.02)", "smooth_eps_0.02", '#B2DDFF'), ("LS(ε=0.05)", "smooth_eps_0.05", '#84CAFF'), ("LS(ε=0.1)", "smooth_eps_0.1", '#2E90FA'), ("LS(ε=0.2)", "smooth_eps_0.2", '#7A5AF8')]
+    op_cfgs = [("Baseline", "baseline", '#F04438'), ("OP(σ=0.005)", "perturbation_0.005", '#A6F4C5'), ("OP(σ=0.01)", "perturbation_0.01", '#6CE9A6'), ("OP(σ=0.015)", "perturbation_0.015", '#32D583'), ("OP(σ=0.02)", "perturbation_0.02", '#12B76A'), ("OP(σ=0.025)", "perturbation_0.025", '#039855'), ("OP(σ=0.03)", "perturbation_0.03", '#027A48')]
 
+    for ax_idx, (ax, cfgs, title) in enumerate([(axes[0], ls_cfgs, 'Label Smoothing (5 models)'), (axes[1], op_cfgs, 'Output Perturbation (7 configs)')]):
+        ax.set_facecolor('white')
+        mx = [max(gm(k, m_key) for _, k, _ in cfgs) for m_key in mk]; mx = [m if m > 0 else 1 for m in mx]
         for nm, ky, cl in cfgs:
-            v = [gm(ky, m_key) / mx[i] for i, m_key in enumerate(mk)]
-            v += [v[0]]  # 闭合
-            lw = 2.8 if ky == 'baseline' else 1.8
-            alpha_fill = 0.10 if ky == 'baseline' else 0.04
-            ax.plot(ag, v, 'o-', lw=lw, label=nm, color=cl, ms=5,
-                    alpha=0.95 if ky == 'baseline' else 0.85)
-            ax.fill(ag, v, alpha=alpha_fill, color=cl)
-
-        ax.set_xticks(ag[:-1])
-        ax.set_xticklabels(ms, fontsize=9, color=COLORS['text'])
-        ax.set_yticklabels([])
-        ax.set_title(title, fontsize=11, fontweight='700',
-                     color=COLORS['text'], pad=18)
-        ax.legend(loc='upper right',
-                  bbox_to_anchor=(1.35 if ax_idx == 1 else 1.30, 1.12),
-                  fontsize=8, framealpha=0.9, edgecolor=COLORS['grid'])
-        ax.spines['polar'].set_color(COLORS['grid'])
-        ax.grid(color=COLORS['grid'], alpha=0.5)
-
+            v = [gm(ky, m_key) / mx[i] for i, m_key in enumerate(mk)]; v += [v[0]]
+            ax.plot(ag, v, 'o-', lw=2.8 if ky == 'baseline' else 1.8, label=nm, color=cl, ms=5, alpha=0.95 if ky == 'baseline' else 0.85)
+            ax.fill(ag, v, alpha=0.10 if ky == 'baseline' else 0.04, color=cl)
+        ax.set_xticks(ag[:-1]); ax.set_xticklabels(ms, fontsize=10, color=COLORS['text']); ax.set_yticklabels([])
+        ax.set_title(title, fontsize=12, fontweight='700', color=COLORS['text'], pad=18)
+        ax.legend(loc='upper right', bbox_to_anchor=(1.35 if ax_idx == 1 else 1.30, 1.12), fontsize=9, framealpha=0.9, edgecolor=COLORS['grid'])
+        ax.spines['polar'].set_color(COLORS['grid']); ax.grid(color=COLORS['grid'], alpha=0.5)
     plt.tight_layout()
     return fig
 
@@ -307,7 +244,7 @@ def fig_perturb_dist():
         ax.hist(mp, bins=bins, alpha=0.6, color=COLORS['accent'], label='Mem+noise', density=True, edgecolor='white')
         ax.hist(np_, bins=bins, alpha=0.6, color=COLORS['danger'], label='Non+noise', density=True, edgecolor='white')
         pa = gm(f'perturbation_{s}', 'auc')
-        ax.set_title(r'OP($\sigma$={})'.format(s) + f'\nAUC={pa:.4f}', fontsize=11, fontweight='semibold'); ax.set_xlabel('Loss', fontsize=10)
+        ax.set_title(f'OP(σ={s})\nAUC={pa:.4f}', fontsize=11, fontweight='semibold'); ax.set_xlabel('Loss', fontsize=10)
         ax.legend(fontsize=9, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
     plt.tight_layout(); return fig
 
@@ -327,7 +264,7 @@ def fig_roc_curves():
         fpr, tpr, _ = roc_curve(y_true, y_scores); ax.plot(fpr, tpr, color=COLORS['danger'], lw=2.5, label=f'Baseline (AUC={bl_auc:.4f})')
         for i, s in enumerate(OP_SIGMAS):
             rng_m = np.random.RandomState(42); rng_nm = np.random.RandomState(137); mp = ml_base + rng_m.normal(0, s, len(ml_base)); np_ = nl_base + rng_nm.normal(0, s, len(nl_base)); y_scores_p = np.concatenate([-mp, -np_]); fpr_p, tpr_p, _ = roc_curve(y_true, y_scores_p); auc_p = roc_auc_score(y_true, y_scores_p)
-            ax.plot(fpr_p, tpr_p, color=COLORS['op_colors'][i], lw=2, label=r'OP($\sigma$={}) (AUC={:.4f})'.format(s, auc_p))
+            ax.plot(fpr_p, tpr_p, color=COLORS['op_colors'][i], lw=2, label=f'OP(σ={s}) (AUC={auc_p:.4f})')
     ax.plot([0,1], [0,1], '--', color=COLORS['text_dim'], lw=1.5, label='Random'); ax.set_xlabel('False Positive Rate', fontsize=12, fontweight='medium'); ax.set_ylabel('True Positive Rate', fontsize=12, fontweight='medium'); ax.set_title('ROC Curves: Output Perturbation', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], loc='lower right'); plt.tight_layout()
     return fig
 
@@ -367,9 +304,9 @@ def fig_tradeoff():
 def fig_auc_trend():
     fig, axes = plt.subplots(1, 2, figsize=(16, 6.5)); apply_light_style(fig, axes); ax = axes[0]; eps_vals = [0.0, 0.02, 0.05, 0.1, 0.2]; auc_vals = [gm(k, 'auc') for k in LS_KEYS]; acc_vals = [gu(k) for k in LS_KEYS]
     ax2 = ax.twinx(); line1 = ax.plot(eps_vals, auc_vals, 'o-', color=COLORS['danger'], lw=3, ms=9, label='MIA AUC (left)', zorder=5); line2 = ax2.plot(eps_vals, acc_vals, 's--', color=COLORS['accent'], lw=3, ms=9, label='Utility % (right)', zorder=5); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5)
-    ax.set_xlabel(r'Label Smoothing $\epsilon$', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium', color=COLORS['danger']); ax2.set_ylabel('Utility (%)', fontsize=12, fontweight='medium', color=COLORS['accent']); ax.set_title('Label Smoothing Trends', fontsize=14, fontweight='bold', pad=15); ax.tick_params(axis='y', labelcolor=COLORS['danger']); ax2.tick_params(axis='y', labelcolor=COLORS['accent']); ax2.spines['right'].set_color(COLORS['accent']); ax2.spines['left'].set_color(COLORS['danger']); lines = line1 + line2; labels = [l.get_label() for l in lines]; ax.legend(lines, labels, fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
+    ax.set_xlabel('Label Smoothing ε', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium', color=COLORS['danger']); ax2.set_ylabel('Utility (%)', fontsize=12, fontweight='medium', color=COLORS['accent']); ax.set_title('Label Smoothing Trends', fontsize=14, fontweight='bold', pad=15); ax.tick_params(axis='y', labelcolor=COLORS['danger']); ax2.tick_params(axis='y', labelcolor=COLORS['accent']); ax2.spines['right'].set_color(COLORS['accent']); ax2.spines['left'].set_color(COLORS['danger']); lines = line1 + line2; labels = [l.get_label() for l in lines]; ax.legend(lines, labels, fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
     ax = axes[1]; sig_vals = OP_SIGMAS; auc_op = [gm(k, 'auc') for k in OP_KEYS]; ax.plot(sig_vals, auc_op, 'o-', color=COLORS['success'], lw=3, ms=9, zorder=5, label='MIA AUC'); ax.axhline(bl_auc, color=COLORS['danger'], ls='--', lw=2, alpha=0.6, label=f'Baseline ({bl_auc:.4f})'); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5, label='Random (0.5)'); ax.fill_between(sig_vals, auc_op, bl_auc, alpha=0.2, color=COLORS['success'], label='AUC Reduction')
-    ax.set_xlabel(r'Perturbation $\sigma$', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium'); ax.set_title('Output Perturbation Trends', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
+    ax.set_xlabel('Perturbation σ', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium'); ax.set_title('Output Perturbation Trends', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
     return fig
 
 def fig_loss_gap_waterfall():
@@ -403,7 +340,6 @@ def cb_sample(src):
     """
     return md, clean_text(s.get('question', '')), clean_text(s.get('answer', ''))
 
-# 🌟 下拉框选项也全部替换为 ε 和 σ
 ATK_CHOICES = (
     ["基线模型 (Baseline)"] +
     [f"标签平滑 (ε={e})" for e in [0.02, 0.05, 0.1, 0.2]] +
@@ -786,16 +722,19 @@ with gr.Blocks(title="MIA攻防研究") as demo:
 
 ---
 """
+            # 🌟🌟🌟 这里就是修复缺失表格行的核心位置！ 🌟🌟🌟
             for sigma in OP_SIGMAS:
                 k = f"perturbation_{sigma}"
                 detail_md += f"""\
 ### 输出扰动 OP(σ={sigma})
 | 指标 | 值 | vs基线 | 变化 |
 |---|---|---|---|
-| AUC | {gm(k,'auc'):.4f} | {bl_auc:.4f} | {gm(k,'auc')-bl_auc:+.4f} |
+| AUC | {gm(k,'auc'):.4f} | {bl_auc:.4f} | {gm(k,'auc')-bl_auc:+.4f} ({(gm(k,'auc')-bl_auc)/bl_auc*100:+.1f}%) |
 | 攻击准确率 | {gm(k,'attack_accuracy'):.4f} | {gm('baseline','attack_accuracy'):.4f} | {gm(k,'attack_accuracy')-gm('baseline','attack_accuracy'):+.4f} |
 | F1 | {gm(k,'f1'):.4f} | {gm('baseline','f1'):.4f} | {gm(k,'f1')-gm('baseline','f1'):+.4f} |
 | TPR@5%FPR | {gm(k,'tpr_at_5fpr'):.4f} | {gm('baseline','tpr_at_5fpr'):.4f} | {gm(k,'tpr_at_5fpr')-gm('baseline','tpr_at_5fpr'):+.4f} |
+| TPR@1%FPR | {gm(k,'tpr_at_1fpr'):.4f} | {gm('baseline','tpr_at_1fpr'):.4f} | {gm(k,'tpr_at_1fpr')-gm('baseline','tpr_at_1fpr'):+.4f} |
+| Loss差距 | {gm(k,'loss_gap'):.4f} | {gm('baseline','loss_gap'):.4f} | {gm(k,'loss_gap')-gm('baseline','loss_gap'):+.4f} |
 | 效用 | {bl_acc:.1f}% | {bl_acc:.1f}% | 0.0% (零损失) |
 
 ---
@@ -851,7 +790,6 @@ with gr.Blocks(title="MIA攻防研究") as demo:
 | σ=0.025 | {gm('perturbation_0.025','auc'):.4f} | {bl_auc-gm('perturbation_0.025','auc'):.4f} | {bl_acc:.1f}% |
 | σ=0.03 | {gm('perturbation_0.03','auc'):.4f} | {bl_auc-gm('perturbation_0.03','auc'):.4f} | {bl_acc:.1f}% |
 
-
 ### 结论四：最佳实践建议
 
 > **推荐组合方案： LS(ε=0.1) + OP(σ=0.02)**