为什么在Scala伴生对象中，隐式对象的优先级高于隐式val？

我一直在玩这个小例子，试图更好地了解scala是如何解析隐式的。我刚刚发现了在伴随对象中声明隐式对象和隐式val之间的区别

示例中名为“证据[-B]”的类型类是逆变的；e、 g.在需要[衍生]证据的情况下，可以使用证据[基础]

我已尝试使用Scala 2.11.7编译以下代码：

trait Evidence[-B] {
  def hello(input: B): String
}

trait Base
object Base {
  implicit object BaseEvidence extends Evidence[Base] {
    override def hello(input: Base): String = "Evidence[Base] from implicit object"
  }
}

class Derived extends Base
object Derived {
  implicit val derivedEvidence: Evidence[Derived] = new Evidence[Derived] {
    override def hello(input: Derived): String = "Evidence[Derived] from implicit val"
  }
}

object Main extends App {
  def hello[A](input: A)(implicit ev: Evidence[A]): Unit = {
    println(ev.hello(input))
  }
  hello(new Derived)
}

但是，编译器给出了一个错误“不明确的隐式值”：

根据我对隐式解析规则的理解，应该首先搜索

派生的
的伴随对象；然后是基类
base
的伴生对象

将隐式val derivedEvidence更改为隐式对象时，代码编译：

object Derived {
  implicit object DerivedEvidence extends Evidence[Derived] {
    override def hello(input: Derived): String = "Evidence[Derived] from implicit object"
}

这种行为的原因是什么

object Derived {
  implicit object DerivedEvidence extends Evidence[Derived] {
    override def hello(input: Derived): String = "Evidence[Derived] from implicit object"
}