Retrieval coverage limits LLM rerankers in cold-start recommendation; a learned hybrid fusion improves pool quality but LLM reranking often degrades end-to-end performance while simpler rankers exploit the pool.
Tallrec: An effective and efficient tuning framework to align large language model with recommendation
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
DINOSAUR augments ANN indices with sampled embeddings to marginalize uncertainty, recovering standard retrieval at zero uncertainty while expanding coverage for uncertain items.
HSTU-based generative recommenders with 1.5 trillion parameters scale as a power law with compute up to GPT-3 scale, outperform baselines by up to 65.8% NDCG, run 5-15x faster than FlashAttention2 on long sequences, and improve online A/B metrics by 12.4%.
A framework for real-time LLM-based user interest personas in large-scale video recommendations, using distillation, async inference, and video clustering to balance interests with novel topics and improve viewer value via A/B tests.
AI brand mentions in ChatGPT/Claude/Gemini conversations causally raise open-web searches and visits by 4.3/2.4/1.0pp via pre-trend event study, stance classifier, and same-category controls on joined clickstream-conversation panel data.
LASAR uses two-stage supervised training plus reinforcement learning to ground semantic IDs, align latent reasoning trajectories to CoT hidden states via KL divergence, and adaptively choose reasoning depth, halving average steps while improving quality on three datasets.
ToolRec introduces dual-level calibration of click data and weighted KTO alignment to improve tool-invoking query recommendations in on-device assistants, reporting CTR gains in large-scale A/B tests.
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.
Greedy linear models without exploration consistently achieve top-tier performance in over 90% of offline dataset evaluations for linear bandit recommenders, with hyperparameter tuning favoring minimal exploration and exposing biases in these protocols.
SAILRec uses dual-side semantic alignment and hierarchical attention steering to improve how LLMs incorporate collaborative embeddings for recommendations, outperforming baselines on MovieLens-1M and Amazon-Book datasets.
Gradient-based representations paired with distribution-matching enable efficient curation of small data subsets that improve performance and training efficiency for continually adapting generative recommenders while maintaining robustness to distributional drift.
citing papers explorer
-
Diagnosing and Mitigating Retrieval Bottlenecks in LLM-Based Cold-Start Recommendation
Retrieval coverage limits LLM rerankers in cold-start recommendation; a learned hybrid fusion improves pool quality but LLM reranking often degrades end-to-end performance while simpler rankers exploit the pool.
-
Distributional Approximate Nearest Neighbour Search for Uncertainty-Aware Retrieval
DINOSAUR augments ANN indices with sampled embeddings to marginalize uncertainty, recovering standard retrieval at zero uncertainty while expanding coverage for uncertain items.
-
LLM-Based User Personas for Recommendations at Scale
A framework for real-time LLM-based user interest personas in large-scale video recommendations, using distillation, async inference, and video clustering to balance interests with novel topics and improve viewer value via A/B tests.
-
LASAR: Latent Adaptive Semantic Aligned Reasoning for Generative Recommendation
LASAR uses two-stage supervised training plus reinforcement learning to ground semantic IDs, align latent reasoning trajectories to CoT hidden states via KL divergence, and adaptively choose reasoning depth, halving average steps while improving quality on three datasets.
-
ToolRec: Calibrated Preference Alignment for Query Recommendation in On-Device Assistants
ToolRec introduces dual-level calibration of click data and weighted KTO alignment to improve tool-invoking query recommendations in on-device assistants, reporting CTR gains in large-scale A/B tests.
-
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.
-
SAILRec: Steering LLM Attention to Dual-Side Semantically Aligned Collaborative Embeddings for Recommendation
SAILRec uses dual-side semantic alignment and hierarchical attention steering to improve how LLMs incorporate collaborative embeddings for recommendations, outperforming baselines on MovieLens-1M and Amazon-Book datasets.
-
Efficient Dataset Selection for Continual Adaptation of Generative Recommenders
Gradient-based representations paired with distribution-matching enable efficient curation of small data subsets that improve performance and training efficiency for continually adapting generative recommenders while maintaining robustness to distributional drift.