Within the context of Chinese university education, effective communication in the field of second language writing heavily relies on lexical complexity, yet the role of writing feedback perception in relation to lexical complexity remains elusive. This study introduces a comprehensive writing feedback perception model encompassing perceptions of teacher-, peer-, and automated written corrective (AWE) feedback, alongside two lexical complexity metrics--Uber and Lambda. By employing latent profile analysis, this research profiles Chinese university students based on their writing feedback perceptions and investigates the resultant lexical complexity variations. Analyzing data from 442 participants, three distinct profiles emerged: students demonstrating preference for feedback from all three agents (teachers, peers, and AWE); students with hold preferable perceptions of teacher and AWE feedback; and students exhibiting no preferable perceptions of feedback of any agents. The means of lexical complexity scores differed significantly across the three profiles. Retrodictive qualitative modeling further unveiled the interplay of feedback perceptions, positive AWE attitudes, and language proficiency in shaping lexical complexity. Remarkably, diverse directional influences emerged across the profiles. Our study underscores the intricate dynamics between writing feedback perception and lexical complexity, with implications for enhancing both teacher feedback literacy and students' feedback perceptions.