Meituan Expands Training Signals for Generative Recommendation
_reachsumit · x · 2026-07-15
Meituan proposed a new training scheme for generative recommendation that no longer relies solely on next-token prediction but incorporates two types of auxiliary supervised signals:
- Temporal supervision: Enhances the modeling of behavioral chronology
- Cross-domain supervision: Integrates information from different domains into the training
The author's core argument is that if generative recommendation only performs next-token prediction, the coverage of training signals will be too narrow; by expanding the supervision objectives, the model can learn richer user behavior patterns and recommendation relationships.
More from Research
- Project APE finds verifier reliability drops when papers contain multiple errors — soumitrashukla9 · 2026-07-22
- Project APE says verifier costs fell about 90x in a year as Chinese open models lead — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Project APE builds its verifier benchmark from 100 AI-written papers with injected errors — soumitrashukla9 · 2026-07-22
- Paper proposes a CRED taxonomy and benchmark to measure research-error detectors — soumitrashukla9 · 2026-07-22
- OpenAI says long-horizon models need safety and alignment checks across full action sequences — rhiever · 2026-07-22