AI search agents learn from mid-search steps
A new method teaches AI search agents which intermediate steps actually helped, without needing a separate reward model.
A new method teaches AI search agents which intermediate steps actually helped, without needing a separate reward model.