Pith. sign in

REVIEW 18 cited by

HLLM: Enhancing Sequential Recommendations via Hierarchical Large Language Models for Item and User Modeling

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2409.12740 v1 pith:342G543T submitted 2024-09-19 cs.IR cs.AI

classification cs.IRcs.AI
keywords hllmrecommendationitemllmsmodelslanguagelargesystems
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large Language Models (LLMs) have achieved remarkable success in various fields, prompting several studies to explore their potential in recommendation systems. However, these attempts have so far resulted in only modest improvements over traditional recommendation models. Moreover, three critical questions remain under-explored: firstly, the real value of LLMs' pre-trained weights, often considered to encapsulate world knowledge; secondly, the necessity of fine-tuning for recommendation tasks; lastly, whether LLMs can exhibit the same scalability benefits in recommendation systems as they do in other domains. In this paper, we propose a novel Hierarchical Large Language Model (HLLM) architecture designed to enhance sequential recommendation systems. Our approach employs a two-tier model: the first Item LLM extracts rich content features from the detailed text description of the item, while the second User LLM utilizes these features to predict users' future interests based on their interaction history. Extensive experiments demonstrate that our method effectively leverages the pre-trained capabilities of open-source LLMs, and further fine-tuning leads to significant performance boosts. Additionally, HLLM achieves excellent scalability, with the largest configuration utilizing 7B parameters for both item feature extraction and user interest modeling. Moreover, HLLM offers excellent training and serving efficiency, making it practical in real-world applications. Evaluations on two large-scale datasets, PixelRec and Amazon Reviews, show that HLLM achieves state-of-the-art results, outperforming traditional ID-based models by a wide margin. In online A/B testing, HLLM showcases notable gains, validating its practical impact in real-world recommendation scenarios. Codes are available at https://github.com/bytedance/HLLM.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 18 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. GALA: Generative Aligned Learning for Adaptive Multimodal Representation in the Taobao Shangou Recommender System

    cs.IR 2026-07 conditional novelty 6.0 of 10

    A three-stage multimodal recommender pipeline with GRPO-based behavior alignment and adaptive ID-content fusion claims a 0.55% online order-volume increase and small offline AUC gains at Taobao Shangou.

  2. From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation

    cs.IR 2026-07 conditional novelty 6.0 of 10

    A feedback-grounded framework discovers recommendation policies by their measured advantage over intent-only baselines and distills them into two latent tokens of a lightweight Semantic-ID recommender.

  3. SpecFormer: Mitigating Embedding and Attention Collapse via Spectral-Aware Transformer for Recommendation

    cs.IR 2026-07 conditional novelty 6.0 of 10

    SpecFormer is a spectral-aware Transformer that flattens the singular-value spectrum of embeddings to prevent embedding/attention collapse, outperforming baselines on CTR benchmarks and scaling with layer depth.

  4. Deep Interest Mining for Intent-Enriched Semantic IDs in Multimodal Generative Recommendation

    cs.IR 2026-03 unverdicted novelty 6.0 of 10

    A new framework integrating deep interest mining, cross-modal semantic alignment, and quality-aware reinforcement learning generates higher-quality Semantic IDs and outperforms prior methods on recommendation benchmarks.

  5. From Scaling to Structured Expressivity: Rethinking Transformers for CTR Prediction

    cs.IR 2025-11 reject novelty 6.0 of 10

    FAT specializes attention by semantic field and reports +0.51% AUC over baselines on Taobao data, but its power-law scaling law is an empirical fit, not a derived prediction.

  6. Time to Split: Exploring Data Splitting Strategies for Offline Evaluation of Sequential Recommenders

    cs.IR 2025-07 conditional novelty 6.0 of 10

    Global temporal splits with Last or Random target selection correlate strongly with realistic successive evaluation, while leave-one-out splits produce inconsistent model rankings across datasets.

  7. VRAgent-R1: Boosting Video Recommendation with MLLM-based Agents via Reinforcement Learning

    cs.MM 2025-07 conditional novelty 6.0 of 10

    VRAgent-R1 uses an MLLM agent to summarize videos and a reinforcement-learned agent to simulate user choices, improving video recommendation and user-decision simulation on MicroLens-100K.

  8. ImpReSS: Implicit Recommender System for Support Conversations

    cs.AI 2025-06 conditional novelty 6.0 of 10

    ImpReSS summarizes support conversations with an LLM, retrieves solution product categories from catalog indexes, and ranks them, reporting MRR@1 of 0.72 to 0.85 across three datasets.

  9. GORACS: Group-level Optimal Transport-guided Coreset Selection for LLM-based Recommender Systems

    cs.IR 2025-06 conditional novelty 6.0 of 10

    GORACS selects small groups of fine-tuning examples via an optimal-transport and gradient-norm proxy objective, outperforming prior coreset methods for LLM-based recommendation.

  10. Generate, Not Recommend: Personalized Multimodal Content Generation

    cs.AI 2025-06 reject novelty 6.0 of 10

    A large multimodal model fine-tuned with reinforcement learning can generate new movie posters and video covers tailored to a user's history, scoring higher on similarity metrics than retrieved items.

  11. SessionRec: Next Session Prediction Paradigm For Generative Sequential Recommendation

    cs.IR 2025-02 conditional novelty 6.0 of 10

    SessionRec redefines generative sequential recommendation as next-session prediction and reports large Recall@500 gains over next-item baselines on two industrial datasets.

  12. Building a User Foundation Model for the Open Web

    cs.LG 2026-07 conditional novelty 5.0 of 10

    A self-supervised Transformer on short open-web browsing sequences improves production CTR and win-rate models and delivers +2.13% live CTR under RTB latency and privacy constraints.

  13. IMFuse: Instance-Aware Multi-Layer Fusion for LLM-Enhanced Sequential Recommendation

    cs.IR 2026-07 conditional novelty 5.0 of 10

    Instance-aware multi-layer fusion of frozen LLM item embeddings improves sequential recommenders by ~6.7% relative over final-layer and generic multi-layer baselines.

  14. TSGR: Taobao Search Generative Retrieval

    cs.IR 2026-07 conditional novelty 5.0 of 10

    TSGR makes Taobao search retrieval value-aware by ordering item IDs by clicks and query intent and re-ranking generated candidates with a value module, improving HR@1000 and GMV.

  15. UniSGR: Unified Framework for Semantic ID Generation and Ranking

    cs.IR 2026-07 conditional novelty 5.0 of 10

    A two-stage generative recommender jointly optimizes value-aware multi-token prediction and multi-objective ranking on shared semantic IDs, with STARK speeding beam search.

  16. Optimizing Recall or Relevance? A Multi-Task Multi-Head Approach for Item-to-Item Retrieval in Recommendation

    cs.IR 2025-06 conditional novelty 5.0 of 10

    A multi-task, multi-head item-to-item retrieval system that merges co-engagement candidates with semantically relevant candidates achieves both higher recall and higher semantic relevance than prior models.

  17. A Contextual-Aware Position Encoding for Sequential Recommendation

    cs.IR 2025-02 conditional novelty 5.0 of 10

    CAPE computes context-dependent positions from item dissimilarity and fuses them into attention, improving sequential recommenders on benchmark and live-traffic tests.

  18. GR-LLMs: Recent Advances in Generative Recommendation Based on Large Language Models

    cs.IR 2025-07 unverdicted novelty 3.0 of 10

    A survey of LLM-based generative recommendation systems, covering application settings, training pipelines, industrial deployment challenges, and future directions.

Pith tools