Item-side structural recommenders outperform user-personalization methods on Moltbook because LLM agents produce stationary, structure-driven engagement rather than learnable preferences.
Generative reasoning re-ranker
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
abstract
Recent studies increasingly explore Large Language Models (LLMs) as a new paradigm for recommendation systems due to their scalability and world knowledge. However, existing work has three key limitations: (1) most efforts focus on retrieval and ranking, while the reranking phase, critical for refining final recommendations, is largely overlooked; (2) LLMs are typically used in zero-shot or supervised fine-tuning settings, leaving their reasoning abilities, especially those enhanced through reinforcement learning (RL) and high-quality reasoning data, underexploited; (3) items are commonly represented by non-semantic IDs, creating major scalability challenges in industrial systems with billions of identifiers. To address these gaps, we propose the Generative Reasoning Reranker (GR2), an end-to-end framework with a three-stage training pipeline tailored for reranking. First, a pretrained LLM is mid-trained on semantic IDs encoded from non-semantic IDs via a tokenizer achieving $\ge$99% uniqueness. Next, a stronger larger-scale LLM generates high-quality reasoning traces through carefully designed prompting and rejection sampling, which are used for supervised fine-tuning to impart foundational reasoning skills. Finally, we apply Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO), enabling scalable RL supervision with verifiable rewards designed specifically for reranking. Experiments on two real-world datasets demonstrate GR2's effectiveness: it surpasses the state-of-the-art OneRec-Think by 2.4% in Recall@5 and 1.3% in NDCG@5. Ablations confirm that advanced reasoning traces yield substantial gains across metrics. We further find that RL reward design is crucial in reranking: LLMs tend to exploit reward hacking by preserving item order, motivating conditional verifiable rewards to mitigate this behavior and optimize reranking performance.
citation-role summary
background 1
citation-polarity summary
fields
cs.IR 9years
2026 9roles
background 1polarities
background 1representative citing papers
PrefixMem encoder for Semantic IDs improves deepest-level accuracy by up to 46% relative and full-SID retrieval recall by up to 22% relative on Pinterest data across LLM families.
CFT + on-policy distillation + RL converts an AR reasoning re-ranker into a block-diffusion model that recovers near-AR accuracy at 2.4–3.5× decode throughput on Amazon Beauty.
GenPage is a transformer that autoregressively generates entire structured Netflix homepages from user prompts, delivering +0.24% engagement lift and 20% latency reduction versus production baseline in online tests.
L2A trains one LLM with input-and-budget-conditioned gates to adapt sparsity across layers, heads, and tokens, tracing the compute-accuracy frontier while staying within 0.6% of dense performance at 34% layer sparsity on tested models.
R2LM combines causal attention with a reverse Mamba SSM sidecar to supply right-side context in dLLMs, claiming 2.4x-12.9x throughput gains over bidirectional dLLMs and 1.9x-2.9x over AR baselines while matching or exceeding quality.
A Llama-based model trained on serialized user stories unifies item, carousel, and search ranking and outperforms specialist baselines offline while improving some online metrics and reducing latency.
GR2 applies mid-training on semantic IDs, reasoning distillation, RL with conditional verifiable rewards, and a context compressor to re-ranking in industrial recsys, reporting +18.7% R@1 over baselines.
TwiSTAR learns to switch between fast SID retrieval and slow rationale-generating reasoning in generative recommendation, yielding better accuracy-latency trade-offs on three datasets.
citing papers explorer
-
Do Recommendation Algorithms Work When Users Are LLM Agents? A Case Study on Moltbook
Item-side structural recommenders outperform user-personalization methods on Moltbook because LLM agents produce stationary, structure-driven engagement rather than learnable preferences.
-
LLMs Need Encoders for Semantic IDs Too
PrefixMem encoder for Semantic IDs improves deepest-level accuracy by up to 46% relative and full-SID retrieval recall by up to 22% relative on Pinterest data across LLM families.
-
Diffusion-GR2: Diffusion Generative Reasoning Re-ranker
CFT + on-policy distillation + RL converts an AR reasoning re-ranker into a block-diffusion model that recovers near-AR accuracy at 2.4–3.5× decode throughput on Amazon Beauty.
-
GenPage: Towards End-to-End Generative Homepage Construction at Netflix
GenPage is a transformer that autoregressively generates entire structured Netflix homepages from user prompts, delivering +0.24% engagement lift and 20% latency reduction versus production baseline in online tests.
-
End-to-End Dynamic Sparsity for Resource-Adaptive LLM Inference
L2A trains one LLM with input-and-budget-conditioned gates to adapt sparsity across layers, heads, and tokens, tracing the compute-accuracy frontier while staying within 0.6% of dense performance at 34% layer sparsity on tested models.
-
Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation
R2LM combines causal attention with a reverse Mamba SSM sidecar to supply right-side context in dLLMs, claiming 2.4x-12.9x throughput gains over bidirectional dLLMs and 1.9x-2.9x over AR baselines while matching or exceeding quality.
-
TubiFM: Unified Item, Carousel, and Search Ranking for Streaming Discovery
A Llama-based model trained on serialized user stories unifies item, carousel, and search ranking and outperforms specialist baselines offline while improving some online metrics and reducing latency.
-
GR2 Technical Report
GR2 applies mid-training on semantic IDs, reasoning distillation, RL with conditional verifiable rewards, and a context compressor to re-ranking in industrial recsys, reporting +18.7% R@1 over baselines.
-
TwiSTAR:Think Fast, Think Slow, Then Act,Generative Recommendation with Adaptive Reasoning
TwiSTAR learns to switch between fast SID retrieval and slow rationale-generating reasoning in generative recommendation, yielding better accuracy-latency trade-offs on three datasets.