Research showing the effectiveness of practice in a controlled environment struggles to confirm this effectiveness on a large scale. Our goal is to solve this scaling problem with a problem posing and solving sequence (18 sessions over 6 weeks), in CE1 and CE2 (2nd and 3rd primary). This sequence concretizes the knowledge of research on learning in problem posing and solving, and the ways to bridge the inequalities between students. The process is carried out in the following manner. 1. Determining what works in the sequence and documenting it. This sequence is implemented in its initial form (S1), in a controlled environment, in the first year of research, by a group of teachers (G1, 8 classes) who are part of the research team, based on an initial M1 model of effectiveness. A method for analyzing the work done by teachers and students is designed and implemented during this first year. It is based on a combination of two types of analysis: on the one hand, statistical analyses to identify what is effective in these practices on the basis of student performance (Evidence Based Practice, EBP); on the other hand, a joint documented analysis of these practices by the members of the research team, bringing together the teachers from the G1 group and the researchers, which determines the reasons for this effectiveness, on the basis of hypermedia systems built from films of the implementation of these practices (Practice Based Evidence, PBE). 2. Validating the effectiveness of the sequence and its accompaniment on a large scale. This method of analysis (combination of EPB/PBE) makes it possible to develop an M2 model of effectiveness, an improved S2 sequence, and a hypermedia system for accompanying the practice. This sequence, the model it embodies, and its accompanying system are tested in the second year of research with a randomized controlled trial on three groups. Two groups implement the sequence: G2 (50 classes) benefiting from the hypermedia support, and G3 (50 classes) not benefiting from this support. One group, G4 (50 classes), works in its usual way on the same mathematical objectives. Our hypothesis is that the students of G2, benefiting from the hypermedia support based on the M2 model, will perform significantly better than the students of G3 and G4. Indeed, the joint work of determining the reasons for the effectiveness of the sequence allows the integration of these reasons in the teaching of the G2 teachers, thanks to the specific support they will have received. The effectiveness observed in a controlled environment is thus attested on a large scale. At the end of the second year, a further improvement of the sequence and the model it embodies produces a sequence S3 and a model M3, thus building an iterative structure of improvement. 3. Refining the sequence and its efficiency model. The third and final year of research is devoted, with a G1' group (16 classes) composed of the G1 group and teachers from the G2 group whose practice will have been shown to be effective, to the refinement of the sequence and its model of effectiveness, in the improvement of the method of analysis developed. The research thus implements a double methodological innovation to solve the problem of scaling up in education: 1) a method of joint analysis of the evolution of the students' performances and of the implementations in the classrooms; 2) a method of documentation of the practices proved to be effective in the elaboration of specific hypermedia systems, intended to instrument the teachers' accompaniment. The dissemination of the results will take place in two directions: towards the research, with two workshops organized in the middle and at the end of the research; towards the teachers - a conference "Problem Posing and Solving" at the end of the research.