Comprehending academic texts requires processing and understanding of complex information. Collaboration may be helpful to support this process. In this work, we investigate whether collaboration can be further supported by a manually predefined concept map that builds up gradually while reading a scientific research article in a digital collaboration environment. We present a randomized controlled trial that compares two conditions: In the intervention (n=50) and control (n=44) group, randomly matched pairs of university students read a scientific text, took notes, discussed the scientific text in real-time in a chat, and summarized the content of the scientific text. The intervention group additionally received the gradually built-up concept map while reading. While the intervention and control groups did not differ in their collaborative knowledge construction or individual knowledge gain, the intervention group was significantly faster at reading the text (F(1,92)=4.70, p=.033) and required fewer notes (F(1,92)=6.66, p=.011) to achieve the same learning outcomes as the control group. Our findings on substantial factual knowledge gained in intervention and control groups underscore the long-standing CSCL tenet that collaboration is an effective learning mode. However, lightweight technological interventions, such as a gradually built-up concept map, can increase learning efficiency in the context of a collaborative understanding of science. The integration of predefined concept maps alongside scientific texts is technically simple. It could, therefore, be incorporated into different digital platforms, supporting collaborative reading of texts and improving the overall learning efficiency of readers.