This paper addresses two persistent challenges in sequential recommendation: (i) evidence insufficiency-cold-start sparsity together with noisy, length-varying item texts; and (ii) opaque modeling of dynamic, multi-faceted intents across long/short horizons. We propose R3-REC (Reasoning-Retrieval-Recommendation), a prompt-centric, retrieval-augmented framework that unifies Multi-level User Intent Reasoning, Item Semantic Extraction, Long-Short Interest Polarity Mining, Similar User Collaborative Enhancement, and Reasoning-based Interest Matching and Scoring. Across ML-1M, Games, and Bundle, R3-REC consistently surpasses strong neural and LLM baselines, yielding improvements up to +10.2% (HR@1) and +6.4% (HR@5) with manageable end-to-end latency. Ablations corroborate complementary gains of all modules.
Figures & tables
Figure 1: Challenges and solution: (i) evidence insufficiency (cold-start + noisy item texts), (ii) opaque modeling of long/short-horizon intents.
Figure 2: Methodology Overview of R3 -REC. The pipeline consists of four stages: (1) Raw Data Processing handles user session streams; (2) Multi-granularity Enhancement distills item semantics via LLMs and models hierarchical user interests (including intent reasoning and long-short term polarity mining); (3) Prompt-guided Assembly Reasoning constructs a comprehensive prompt context by inserting retrieved similar users, inferred intents, and candidates; (4) Context-aware Recommendation employs an LLM to perform feature analysis, generating both ranking scores and interpretative reasoning.
Conventional Methods
Single-Intent Methods
Multi-Intent Methods
LLM Methods
Ours
Improve
p -value
Data
Metrics
MostPop
FPMC
NARM
STAMP
GCE-GNN
MCPRN
HIDE
Atten-Mixer
NIR
PO4ISR
RecMind
UniRec
R 3 -REC
ML-1M
HR@1
0.0004
0.1132
0.1692 ⋆
0.1584
0.1312
0.1434
0.1498
0.1490
0.0572
0.2000
0.2700 ∗
0.2650
0.287
+6.3%
4.3e−3
HR@5
0.0070
0.3748
0.5230 ⋆
0.5078
0.4748
0.4788
0.4998
0.4932
0.2326
0.5510
0.5720
0.5910 ∗
0.597
+1.0%
5.9e−2
HR@10
0.0120
0.5150
0.7050 ⋆
0.6920
0.6600
0.6640
0.6820
0.6760
0.3550
0.7240
0.7480
0.7620 ∗
0.769
+0.9%
4.6e−2
Games
HR@1
–
0.0498
0.0572
0.0556
0.0692
0.0522
0.0696
0.0530
0.1168 ⋆
0.2588
0.3450 ∗
0.3400
0.379
+9.9%
6.4e−5
HR@5
–
0.2564
0.2574
0.2586
0.2744
0.2416
0.2694
0.2472
0.3406 ⋆
0.5866
0.6550 ∗
0.6450
0.679
+3.7%
3.5e−5
Table 1: Main comparison on ML-1M , Games , and Bundle . Higher is better. Best results are in bold ; second-best are marked with ∗ .
Figure 3: Ablation on R 3 -REC (HR@1 / HR@5 / NDCG@5). Lower bars on variants indicate performance drops relative to the full model.
Figure 4: Hyperparameter sensitivity under two factorizations.
Model
Avg. Tokens
Latency
Est. Cost
HR@1
Input
Output
(ms)
($)
(Games)
NIR [ 21 ]
150
10
250
< 1e-4
0.1168
PO4ISR [ 19 ]
350
15
380
2e-4
0.2588
RecMind [ 23 ]
600
80
920
4e-4
0.3450
R 3 -REC
1250
10
580
6e-4
0.3790
Table 2: Efficiency Analysis (20-item slate). Est. Cost : Based on GPT-3.5 pricing. Latency : Ranking phase only.
Large language models (LLMs) have been applied to sequential recommendation by formulating it as a natural language task. Previous work has improved personalization by incorporating collaborative and sequential signals through input conditioning or LLM fine-tuning. However, existing approaches often rely on one or more of the following: LLM fine-tuning, additional architectural modules, representation distillation, or item-level conditioning over long interaction histories, increasing training complexity and deployment cost. We propose REPREC, a lightweight framework that reformulates LLM-based sequential recommendation through lightweight user representation alignment. REPREC maps a fixed-size user embedding from a frozen sequential encoder into a small set of learned soft tokens through a lightweight MLP injector that conditions a frozen LLM, leaving both pretrained backbones unchanged while training only the injector. We conducted exhaustive experiments on multiple benchmark datasets and demonstrate that REPREC consistently outperforms LoRA while remaining compatible with different pretrained sequential encoders and LLM backbones, enabling a modular and production-friendly recommendation pipeline without modifying either pretrained component. The gains are particularly pronounced for casual and core users across all datasets, highlighting REPREC's effectiveness in low-data regimes. Finally, when trained on short prompt histories and evaluated with longer contexts, REPREC maintains 85-100% of LoRA's performance while reducing per-epoch training time by an average of 1.51X, demonstrating an effective balance between recommendation quality and computational efficiency for production deployment. The code is available at https://github.com/phdbotcode/REPREC
Harshini Kavuru, Dwipam Katariya, Giri Iyengar +3
The Ohio State University, USA · Capital One, AI Foundations, USA
Large language models (LLMs) have demonstrated prominent reasoning capabilities in recommendation tasks by transforming them into text-generation tasks. However, existing approaches either disregard or ineffectively model the user-item high-order interactions. To this end, this paper presents an enhanced LLM-based recommender (ELMRec). We enhance whole-word embeddings to substantially enhance LLMs' interpretation of graph-constructed interactions for recommendations, without requiring graph pre-training. This finding may inspire endeavors to incorporate rich knowledge graphs into LLM-based recommenders via whole-word embedding. We also found that LLMs often recommend items based on users' earlier interactions rather than recent ones, and present a reranking solution. Our ELMRec outperforms state-of-the-art (SOTA) methods in both direct and sequential recommendations.
Xinfeng Wang, Jin Cui, Fumiyo Fukumoto +1
Graduate School of Engineering University of Yamanashi, Kofu, Japan · Interdisciplinary Graduate School University of Yamanashi, Kofu, Japan
Sequential recommender systems typically infer user preferences through single-pass encoding of interaction histories without iterative refinement, relying on increasingly deep architectures to capture complex patterns. In this work, we revisit sequential recommendation from a recursive inference perspective: can user preferences be modeled as a persistent latent state that is recursively refined? We propose RecRec (Recursive Recommendation), a lightweight model that maintains a compact latent state and updates it through a shared recursive module conditioned on interaction evidence. Unlike prior recursive models, RecRec introduces an evidence-anchored correction mechanism that stabilizes refinement by grounding each update in the original interaction context, preventing semantic drift during deep recursive reasoning. Experiments on three benchmark datasets under standard evaluation protocols show that RecRec matches or outperforms state-of-the-art sequential, graph-based, and reasoning-enhanced recommenders while using only 3.9M to 14M parameters. Ablation studies demonstrate that both recursive refinement and the evidence-anchored correction gate contribute significantly to performance, highlighting the effectiveness of recursive latent inference as a scalable alternative to deeper or language-based architectures. Code is available at https://anonymous.4open.science/r/RecRec-6B67/README.md.