Personalized LLM Recommendation with Semantic IDs: Autoregressive Beam Search vs. Multi-Token Prediction
The experiments show single-head beam search consistently beats MTP on most product-critical metrics: ranking and content QA, and trains at roughly 2.6x lower dollar cost on Qwen3-8B scale; MTP wins only on raw inference throughput.
Sejoon Oh, Fernando Amat Gil, Evan Smoth-ers et al.
· 0 citations