The patent authority eventually faces either an adverse selection problem or a moral hazard problem that asymmetric information causes when it has imperfect information about the attributes of innovations. This paper analyzes whether the current patent system with uniform patent life can still be an effective incentive scheme of R&D investments in cumulative innovations with inventor privately informed on innovation cost. This paper looks into the nature of sequential innovations and bilateral private information inherent in cumulative innovations, specifically research tools and its commercialization. Under the current patent system, the inventor of a research tool (inventor A) can collect profit either by self-development or by forming an R&D joint venture with the subsequent inventor (inventor B) since revenues accrue only to commercial product. Due to the inventor's private information, however, a negotiation over an R&D joint venture between sequential inventors is not always reached. That leads to patent race between inventors A and B, which results in an inefficient commercialization. Bilateral private information and sequential structure inherent in cumulative innovations call for a new paradigm to design a self-selecting mechanism beyond a general form of 1 principal - 1 agent model.