Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation

Zhanshuo Cao; Jiakai Tang; Yuqi Zhou; Sunhao Dai; Jun Xu

cs.IROct 21, 2026

Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation

Authors: Zhanshuo Cao, Jiakai Tang, Yuqi Zhou, Sunhao Dai, Jun Xu

Organizations: Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China

Abstract

Existing sequential recommenders usually allocate a fixed computational budget to every item in a user's interaction history, regardless of the item's contribution to the final prediction. We challenge this de facto design with two diagnostic findings. (1) Non-uniform computational need: items within the same sequence vary drastically in how much computation they require. Most converge after shallow processing while a critical minority only stabilizes at full depth. (2) Suboptimality of fixed allocation: given oracle access to the target item, a non-uniform depth assignment consistently outranks the fixed one, indicating that existing uniform computation both wastes capacity on already-converged items and sacrifices overall recommendation quality. These observations motivate a shift from fixed-depth computation to item-level adaptive reasoning. To this end, we propose LoopRec. LoopRec recursively applies a single weight-shared block and introduces a lightweight gate at each recursion boundary that makes a binary exit-or-continue decision for every item independently. Items whose representations have sufficiently matured exit early, while those requiring deeper reasoning continue to subsequent recursions. This design converts the fixed global depth into a learned, per-item property without auxiliary depth labels, and is readily adaptable to diverse backbone architectures. Extensive experiments on public benchmarks with four backbone families demonstrate that LoopRec consistently outperforms matched fixed-depth counterparts. Further analysis reveals that LoopRec's learned allocation exhibits interpretable patterns aligned with our post-hoc oracle analysis: lower-frequency items and more recent interactions naturally receive deeper computation. Our code is available at https://github.com/Elpmis117/LoopRec.

Figures & tables

Figure 1. Item saturation depth

\ell_{i}^{\star}

on Office. (a) Item-level distribution across the test set. (b) Per-sequence depth composition. Each violin shows the distribution of within-sequence shares, with dots and bars marking medians and IQRs.

Sequence	Seq. 1	Seq. 2	Seq. 3	Seq. 4	Seq. 5
Rank	$24{\to}13$	$7{\to}4$	$7{\to}5$	$12{\to}10$	$19{\to}18$
Gain	$+11$	$+3$	$+2$	$+2$	$+1$
Rel. tail comp.	$40.0\%$	$66.7\%$	$53.3\%$	$60.0\%$	$60.0\%$

Table 1. Rank improvements and relative tail computation for the sequences in Figure 2 (b), normalized by the fixed-depth allocation.

Figure 2. Restricted tail-5 oracle on Office. (a) Relative gains of the flexible oracle over the fixed-depth baseline. (b) Oracle-selected tail depths for sample sequences. Numbers denote depths, and columns 0–4 are the five most recent positions.

Figure 3. Overview of LoopRec. At each recursion boundary, depth-routing gates decide which positions stop and which continue through the shared block.

Dataset	Users	Items	Interactions	Avg. Len.	Sparsity
Software	13,444	9,706	227,637	16.93	99.83%
Office	11,468	28,279	190,465	16.61	99.94%
Patio	14,454	37,179	229,086	15.85	99.96%

Table 2. Statistics of experimental datasets.

Backbone	Model	Software				Office				Patio
Backbone	Model	M@10	M@20	N@10	N@20	M@10	M@20	N@10	N@20	M@10	M@20	N@10	N@20
SASRec	Vanilla	0.0589	0.0629	0.0796	0.0944	0.0508	0.0513	0.0562	0.0582	0.0394	0.0399	0.0429	0.0448
	+LoopRec	0.0613	0.0659	0.0830	0.0998	0.0523	0.0530	0.0571	0.0595	0.0411	0.0416	0.0441	0.0458
	Improv.	+4.14%*	+4.77%*	+4.27%*	+5.68%*	+3.03%*	+3.24%*	+1.49%*	+2.20%*	+4.42%*	+4.26%*	+2.75%*	+2.19%*
BSARec	Vanilla	0.0607	0.0657	0.0803	0.0987	0.0508	0.0516	0.0565	0.0596	0.0391	0.0398	0.0426	0.0453
	+LoopRec	0.0619	0.0665	0.0828	0.0998	0.0519	0.0527	0.0571	0.0602	0.0415	0.0422	0.0445	0.0469

Table 3. Main results across three Amazon benchmarks, comparing each fixed-depth backbone with its LoopRec counterpart. M@K and N@K denote MRR@K and NDCG@K, and values are test means over five runs. The “*” denotes the statistical significance (

p<0.05

) of the results of LoopRec compared to the corresponding fixed-depth backbone under a paired t-test. Within each backbone block, the better value is shown in bold .

Figure 4. Average selected recursion round across relative-position deciles (D1: earliest, D10: most recent), on Office and Patio under two LxRy configurations.

Train freq.	Tokens	Mean depth	Depth 1	Depth 4
1	505	2.160	50.3%	29.1%
2–4	1,202	2.084	53.7%	27.0%
5–9	1,603	2.135	53.3%	29.3%
10–19	2,247	2.008	60.8%	28.1%
20–49	4,123	1.818	69.2%	24.1%
50–99	3,668	1.577	79.1%	17.7%

Table 4. Selected depth by item training-time frequency on Office. Depth columns report the share of tokens assigned to each depth.

Figure 5. Saturation depth on Office under a three-layer SASRec. (a) Mean depth across relative-position quartiles. (b) Depth shares across training-frequency buckets.

Model	Software		Office		Patio
Model	M@20	N@20	M@20	N@20	M@20	N@20
LoopRec	0.0659	0.0998	0.0530	0.0595	0.0416	0.0458
Seq-level Loop	0.0634	0.0950	0.0525	0.0594	0.0390	0.0447
Vanilla	0.0629	0.0944	0.0513	0.0582	0.0399	0.0448

Table 5. Comparison of sequence-level looping and item-level depth allocation. Vanilla denotes the original SASRec. Results are five-seed test means. M and N denote MRR and NDCG, respectively.

Ablation	Software		Office		Patio
Ablation	M@20	N@20	M@20	N@20	M@20	N@20
LoopRec	0.0659	0.0998	0.0530	0.0595	0.0416	0.0458
w/o warm-up	0.0615	0.0942	0.0520	0.0583	0.0401	0.0450
w/o $\mathcal{L}_{\mathrm{bal}}$	0.0655	0.0995	0.0507	0.0576	0.0385	0.0431
w/o $\mathcal{L}_{z}$	0.0639	0.0973	0.0507	0.0574	0.0387	0.0432
w/o $\mathcal{L}_{\mathrm{bal}}$ & $\mathcal{L}_{z}$	0.0632	0.0963	0.0505	0.0574	0.0386	0.0431

Table 6. Ablation of routing warm-up and router-side regularizers. Results are five-seed test means. M and N denote MRR and NDCG, respectively.

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 6. Selected-depth distributions on Office for four LxRy configurations. The y-axis is the per-sequence share of positions at each effective depth.

Pos.	Item	Category	Freq.	Depth
1	ValueMailers poly mailers	Envelope mailer	56	1
2	Uboxes poly mailers	Envelope mailer	261	1
3	ValueMailers poly mailers	Envelope mailer	517	1
4	EasyPAG desk organizer	Desk organizer	5	1
5	Quartet whiteboard	Dry erase board	259	1
6	Aviditi box mailers	Box mailer	1	2

Appendix

Table 7. Item-level routing of one Office sequence. Freq. is training-time interaction frequency. Depth is the selected depth.

Figure 7. Category-level routed depth versus item popularity on Office. Each point is a category leaf (size

\propto

token count). Blue and orange mark the five highest- and lowest-depth categories, and the solid line shows the overall trend.

Method	MRR@10	MRR@20	NDCG@10	NDCG@20
Vanilla	0.0508	0.0513	0.0562	0.0582
Position	0.0508	0.0515	0.0558	0.0585
Frequency	0.0510	0.0516	0.0561	0.0584
LoopRec	0.0523	0.0530	0.0571	0.0595

Appendix

Table 8. Comparison with heuristic routing on Office using SASRec. Values are test means over five seeds. Bold values denote the highest mean for each metric.

Setting	Ref.	Compute	Ref. ms	LoopRec ms	$\Delta$ ms	Latency
L2R4	SASRec-8L	0.407 $\times$	91.230	79.273	$-11.957$	0.869 $\times$
L2R3	SASRec-6L	0.680 $\times$	69.684	63.196	$-6.488$	0.907 $\times$
L2R2	SASRec-4L	0.755 $\times$	47.624	47.379	$-0.245$	0.995 $\times$

Appendix

Table 9. Compute budget and encoder latency on Office (single A6000 GPU). Compute and Latency are normalized by the matched fixed-depth reference. Ref. ms and LoopRec ms are median encoder times.

Sequential Recommendation

Sequential recommendation predicts what a user will choose next from ordered interactions, adapting to changing interests and context.

See all 132 papers

RecRec: Recursive Refinement for Sequential Recommendation
Jul 12, 2026Pervez Shaik, Prosenjit Biswas, Abhinav Thorat +2Sequential Recommendation Iterative Refinement
Sony Research India India
REPREC: Representation Driven Parameter-Efficient Recommendation System
Jul 24, 2026Harshini Kavuru, Dwipam Katariya, Giri Iyengar +3Sequential Recommendation Large Language Model-Based Recommendation
The Ohio State University, USA · Capital One, AI Foundations, USA
Rethinking Convolutional Networks for Attribute-Aware Sequential Recommendation
May 6, 2026Shereen Elsayed, Ngoc Son Le, Ahmed Rashed +1Sequential Recommendation Convolutional Neural Networks
Information Systems and Machine Learning Lab (ISMLL), University of Hildesheim, Germany · VWFS Data Analytics Research Center (VWFS DARC), University of Hildesheim, Germany · Volkswagen Financial Services AG, Germany

Jul 12, 2026J/K move · Enter open · S save

RecRec: Recursive Refinement for Sequential Recommendation

Pervez Shaik, Prosenjit Biswas, Abhinav Thorat, Ravi Kolla+1

Sony Research India India

Sequential Recommendation Iterative Refinement User Preference Modeling Sequence Modeling

Open paper AskPDF arXiv

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.

Abstract

Sequential Recommendation

Explore similar work

RecRec: Recursive Refinement for Sequential Recommendation

REPREC: Representation Driven Parameter-Efficient Recommendation System

Rethinking Convolutional Networks for Attribute-Aware Sequential Recommendation