polysemy

DSRec is a novel dual-interest sequential recommendation model using multi-granular State Space Models to capture long-term and short-term user interests, addressing item polysemy and achieving superior performance on benchmarks.