ManCAR: Manifold-Constrained Latent Reasoning with Adaptive Test-Time Computation for Sequential Recommendation

2026-02-23

AI Summary

This work addresses the issue of representation drift in existing multi-step reasoning approaches for sequential recommendation, which often lack explicit constraints on the feasibility of intermediate states, leading latent representations into invalid regions. To mitigate this, the authors model recommendation reasoning as a navigation process on a collaborative manifold. They construct a local intent prior from the user’s recent interactions and align the predicted distribution with this prior during training to constrain the reasoning trajectory within the valid manifold. At inference time, an adaptive stopping mechanism is introduced to prevent over-optimization. By explicitly incorporating the topological structure of the collaborative manifold into the reasoning process—combining graph-based neighborhood modeling with variational inference—the proposed method achieves significant performance gains, outperforming state-of-the-art approaches across seven benchmark datasets, with up to a 46.88% relative improvement in NDCG@10.

Technology Category

• Reasoning under Uncertainty: Sequential Decision Making
• Knowledge Representation and Reasoning: Preferences
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Sequential recommendation increasingly employs latent multi-step reasoning to enhance test-time computation. Despite empirical gains, existing approaches largely drive intermediate reasoning states via target-dominant objectives without imposing explicit feasibility constraints. This results in latent drift, where reasoning trajectories deviate into implausible regions. We argue that effective recommendation reasoning should instead be viewed as navigation on a collaborative manifold rather than free-form latent refinement. To this end, we propose ManCAR (Manifold-Constrained Adaptive Reasoning), a principled framework that grounds reasoning within the topology of a global interaction graph. ManCAR constructs a local intent prior from the collaborative neighborhood of a user's recent actions, represented as a distribution over the item simplex. During training, the model progressively aligns its latent predictive distribution with this prior, forcing the reasoning trajectory to remain within the valid manifold. At test time, reasoning proceeds adaptively until the predictive distribution stabilizes, avoiding over-refinement. We provide a variational interpretation of ManCAR to theoretically validate its drift-prevention and adaptive test-time stopping mechanisms. Experiments on seven benchmarks demonstrate that ManCAR consistently outperforms state-of-the-art baselines, achieving up to a 46.88% relative improvement w.r.t. NDCG@10. Our code is available at https://github.com/FuCongResearchSquad/ManCAR.

Problem

• sequential recommendation
• latent reasoning
• manifold constraint
• latent drift
• test-time computation

Innovation

• manifold-constrained reasoning
• adaptive test-time computation
• collaborative manifold
• latent drift prevention
• sequential recommendation