LASAR: Latent Adaptive Semantic Aligned Reasoning for Generative Recommendation

2026-05-11

AI Summary

This work addresses the high latency of token-by-token generation in large language models for generative recommendation, as well as key limitations of latent-space reasoning—including semantic misalignment, representation drift, and fixed inference depth—by proposing the LASAR framework. LASAR introduces adaptive latent-space reasoning into generative recommendation through a two-stage training process: first aligning semantic IDs, then performing latent-space reasoning. It integrates explicit chain-of-thought semantic alignment, stepwise bidirectional KL regularization, and a sample-level dynamic inference depth prediction mechanism, trained via an SFT-then-RL paradigm with GRPO/REINFORCE-based policy optimization. Experiments on three real-world datasets demonstrate that LASAR significantly outperforms baseline methods, achieving approximately 20× faster inference than explicit chain-of-thought approaches, nearly halving the average number of latent reasoning steps, and incurring only minimal additional latency overhead.

Abstract

Large Language Models (LLMs) have demonstrated powerful reasoning capabilities through Chain-of-Thought (CoT) in various tasks, yet the inefficiency of token-by-token generation hinders real-world deployment in latency-sensitive recommender systems. Latent reasoning has emerged as an effective paradigm in LLMs, performing multi-step inference in a continuous hidden-state space to achieve stronger reasoning at lower cost. However, this paradigm remains underexplored in mainstream generative recommendation. Adapting it reveals three unique challenges: (1) the gap between prior-less Semantic ID (SID) symbols and continuous latent reasoning - SIDs lack pre-trained semantics, hindering joint optimization; (2) representation drift due to a lack of reasoning chain supervision; and (3) the suboptimality of applying a globally fixed reasoning depth. To address these, we propose LASAR (Latent Adaptive Semantic Aligned Reasoning), an SFT-then-RL framework. First, we bridge this gap via two-stage training: Stage 1 grounds SID semantics before Stage 2 introduces latent reasoning, ensuring efficient convergence. Second, we mitigate representation drift through explicit CoT semantic alignment. Step-wise bidirectional KL divergence constrains the latent reasoning trajectory using hidden-state anchors extracted from CoT text, while a Policy Head predicts per-sample reasoning depth. Third, during the GRPO-based RL phase, terminal-only KL alignment accommodates variable-length reasoning, and REINFORCE optimizes the Policy Head to dynamically allocate steps. This nearly halves the average latent step count while simultaneously improving recommendation quality. Experiments on three real-world datasets demonstrate that LASAR outperforms all baselines. It adds marginal inference latency and is roughly 20 times faster than generating explicit CoT text.

Problem

• latent reasoning
• generative recommendation
• semantic alignment
• representation drift
• reasoning depth

Innovation

• Latent Reasoning
• Semantic Alignment
• Adaptive Reasoning Depth
• Generative Recommendation
• Chain-of-Thought