An arXiv preprint introduces MAVE, a multi-agent framework designed to improve pedagogical feedback in online classes. The paper notes that timely feedback is hard to scale in this setting because course videos are long and noisy, and student responses are difficult to locate within them.
According to the abstract, current automated systems tend to produce generic advice that is only weakly grounded in what actually happens in the classroom. MAVE is positioned as a response to that problem, using a multi-agent setup to generate feedback that is more closely tied to real classroom behavior.
Because the abstract is brief, the paper's specific mechanisms and evaluation results are not described in the announcement. The contribution, as stated, is the framework itself and its aim to ground feedback in video evidence.