SQL - Eliminación avanzada de duplicados usando ID NOT IN

Puedes usar ROW_NUMBER función analítica y correlacionar la eliminación usando el ROWID pseudo-columna:

DELETE FROM my_table
WHERE ROWID IN (
  SELECT ROWID
  FROM   (
    SELECT ROW_NUMBER() OVER (
             PARTITION BY student_id, subject_id, class_id
             ORDER BY expertise_lvl DESC
           ) AS rn
    FROM   my_table
  )
  WHERE  rn > 1
)

db<>fiddle aquí