Abnormal performance of the min-max attack method?

[Gmrider13]

Hi, I run the test_min_max.py and get worse attack performance compared with the paper: in citreseer:

in cora:

performance in paper:

I notice that the author "set 20 steps of inner maximization per iteration" so I just modified the topology_attack.py from：

victim_model.train()
modified_adj = self.get_modified_adj(ori_adj)
adj_norm = utils.normalize_adj_tensor(modified_adj)
output = victim_model(ori_features, adj_norm)
loss = self._loss(output[idx_train], labels[idx_train])

optimizer.zero_grad()
loss.backward()
optimizer.step()

to

victim_model.train()
modified_adj = self.get_modified_adj(ori_adj)
adj_norm = utils.normalize_adj_tensor(modified_adj)
for epoch in range(inner_iter):
    output = victim_model(ori_features, adj_norm)
    loss = self._loss(output[idx_train], labels[idx_train])
    optimizer.zero_grad()
    
    loss.backward()
    optimizer.step()

but achieve higher accuracy in cora and citeseer:

Do you have any idea about that?

Thanks