- Rating
- 2(+1)
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
This paper proposes LoopGR, a SID-based generative recommender that alternates a shared decoder with a recurrent context updater. It also introduces recurrent trajectory self-distillation (RTSD) to transfer a high-budget teacher to a lower-budget student. Experiments on three Amazon Reviews 2014 categories report competitive recommendation accuracy with fewer parameters than an encoder–decoder baseline.
Critical strengths
The paper addresses a relevant trade-off between model size and inference cost. The latency measurements for compressed students and the ablations of different context updaters provide useful empirical insights.
Critical weaknesses
The novelty and scaling claims are not sufficiently established. LoopGR combines shared recurrent decoding with a lightweight context updater, while RTSD aligns intermediate predictions and their changes. The paper does not clearly establish a sufficiently distinct methodological contribution beyond these design choices. Moreover, the evaluation is limited to three Amazon Reviews 2014 categories with short average interaction histories. Parameter and FLOP scaling comparisons are presented only on CDs and Vinyl, which is insufficient to support the broader claim that recurrent depth is a general scaling dimension for generative recommendation.
The experimental protocol is also under-specified: the temporal train/validation/test split, candidate-ranking procedure, and variability across runs are not reported clearly. In addition, Table 2 compares RTSD at half the recurrent budget against alternative objectives at the full budget, making it difficult to isolate the benefit of the proposed distillation objective under matched inference costs.
Questions for authors
What exact temporal split and candidate-ranking protocol are used, and do the reported gains persist across multiple random seeds? Can RTSD be compared with alternative objectives at the same recurrent budget or latency?
5 comments
Response to Reviewer qsMQ (Part 2)
(W3, Q1) Data splitting, candidate ranking and robustness across random seeds.
All baselines use the same data split, and the additional five-seed experiments show that LoopGR's advantages persist in the mean results. We specify the public and industrial evaluation settings below.
-
Public datasets. We use the Beauty, Sports and CDs categories of Amazon Reviews 2014. Following the preprocessing operations in TIGER, we construct item sequences from users' review histories, sort interactions by timestamp, and filter out users with fewer than five reviews. We use a per-user chronological leave-one-out split: the last interaction is the test target, the penultimate interaction is the validation target, and the earlier interactions form the training sequence. LoopGR training examples are constructed with sliding windows over the training sequence, pairing each target with its preceding history. The validation history contains only interactions preceding the validation target. The test history additionally includes the validation interaction. The same split is used across baselines. For LoopGR's candidate generation and ranking on Amazon, we use autoregressive beam search with beam size 20 and Trie-based SID constraints. The Trie restricts generated prefixes to valid catalog SID prefixes. Completed SID sequences are ranked by cumulative token log-probability. Recall/Hit@5, Recall/Hit@10, NDCG@5, and NDCG@10 are computed by matching the held-out item's complete SID against the ranked generated sequences. With one target per example, Recall@K and Hit@K coincide. This evaluation uses generative retrieval over catalog SIDs without constructing a sampled-negative ranking pool.
-
Industrial dataset. We train on millions of samples from the first seven days and evaluate predictions on the eighth day. The historical item sequence length is 100, and each item has four SID tokens, giving an input SID sequence length of 400. The industrial evaluation reports Hit@5, Hit@10, Hit@20, Hit@50 and Hit@100, with results for Lazy Decoder-only, TIGER, BARGE and the four LoopGR variants shown in W2. This chronological setting evaluates recommendation on the subsequent day's interactions. Besides, we use autoregressive beam search with beam widths of 32, 32, 32 and 100 at the four SID decoding positions, respectively.
-
Random-seed evaluation. We conducted five runs with different random seeds for TIGER, ActionPiece and LoopGR on each of the three Amazon datasets. The full mean ± standard deviation results are provided below. LoopGR achieves higher mean performance than TIGER on all 12 metrics and than ActionPiece on 8 of 12 metrics. We will include these protocol details and repeated-run results in the revised manuscript.
| Datasets | Model | R@5 | N@5 | R@10 | N@10 |
|---|---|---|---|---|---|
| Beauty | TIGER | 0.0466±0.0011 | 0.0321±0.0002 | 0.0682±0.0014 | 0.0391±0.0007 |
| ActionPiece | |||||
| LoopGR | 0.0566±0.0002 | 0.0388±0.0008 | 0.0808±0.0006 | 0.0467±0.0007 | |
| Sports and Outdoors | TIGER | 0.0275±0.0011 | 0.0182±0.0005 | 0.0428±0.0028 | 0.0231±0.0006 |
| ActionPiece | 0.0500±0.0007 | ||||
| LoopGR | 0.0334±0.0005 | 0.0225±0.0003 | 0.0276±0.0003 | ||
| CDs and Vinyl | TIGER | 0.0496±0.0004 | 0.0327±0.0002 | 0.0760±0.0012 | 0.0412±0.0003 |
| ActionPiece | 0.0544±0.0005 | 0.0830±0.0008 | 0.0451±0.0005 | ||
| LoopGR | 0.0361±0.0004 |
Response to Reviewer qsMQ (Part 3)
(W4, Q2) RTSD comparisons under the same recurrent inference budget.
The additional experiments compare RTSD with alternative student-training strategies and component ablations using the same 2-loop architecture, directly controlling the recurrent inference budget. We train a 4-loop teacher on each Amazon dataset and evaluate six alternatives to full RTSD: (1) training a 2-loop student from scratch; (2) reducing the frozen teacher's inference budget from four loops to two; (3) initializing the student from the teacher and fine-tuning with only the task loss; (4) output-only distillation; (5) removing the trajectory loss from RTSD; and (6) removing the direction loss from RTSD. The following table reports all four metrics on all three datasets.
The results support the contribution of trajectory supervision beyond teacher initialization, task-only adaptation and output-only distillation at the same student loop budget. Directly truncating the frozen teacher to two loops substantially degrades performance, whereas RTSD retains at least 97% of the teacher's score on every metric while halving the recurrent budget. On Sports, the student also exceeds the teacher on Recall@10 and NDCG@10. The ablation results show that trajectory-distribution alignment improves performance across most metrics. Removing this component reduces performance on 9 of 12 metrics. These results distinguish the roles of the two loss terms and substantiate RTSD's effectiveness under a fixed recurrent inference budget. We will add the full fixed-budget comparison to the revised ablation study.
| Ablation | Beauty R@5 | N@5 | R@10 | N@10 | Sports R@5 | N@5 | R@10 | N@10 | CDs R@5 | N@5 | R@10 | N@10 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LoopGR (Teacher, 4 loops) | 0.0546 | 0.0373 | 0.0806 | 0.0456 | 0.0330 | 0.0222 | 0.0537 | 0.0360 | 0.0791 | 0.0442 | ||
| Train 2-loop student from scratch | 0.0517 | 0.0350 | 0.0754 | 0.0426 | 0.0218 | 0.0270 | 0.0507 | 0.0338 | 0.0767 | 0.0422 | ||
| Reduce teacher inference budget (4 to 2 loops) | 0.0276 | 0.0183 | 0.0444 | 0.0237 | 0.0199 | 0.0127 | 0.0316 | 0.0165 | 0.0239 | 0.0156 | 0.0377 | 0.0201 |
| Fine-tune 2-loop student from 4-loop teacher | 0.0515 | 0.0351 | 0.0779 | 0.0436 | 0.0311 | 0.0212 | 0.0467 | 0.0262 | 0.0512 | 0.0346 | 0.0778 | 0.0431 |
| output-only distillation | 0.0530 | 0.0366 | 0.0777 | 0.0446 | 0.0316 | 0.0210 | 0.0474 | 0.0261 | 0.0514 | 0.0345 | 0.0762 | 0.0425 |
| RTSD (Full, 2 loops) | 0.0451 | 0.0324 | 0.0495 | 0.0274 | 0.0781 | 0.0436 | ||||||
| RTSD w/o trajectory loss | 0.0521 | 0.0360 | 0.0447 | 0.0316 | 0.0215 | 0.0474 | 0.0266 | 0.0518 | 0.0350 | 0.0436 | ||
| RTSD w/o direction loss | 0.0533 | 0.0373 | 0.0323 | 0.0485 | 0.0518 | 0.0352 | 0.0785 |
Response to Reviewer qsMQ (Part 1)
We thank the reviewer for the advice.
(W1) The methodological contributions of LoopGR and RTSD.
LoopGR introduces an encoder–updater–decoder architecture organized through hierarchical recurrent depth, together with a training strategy that separates trajectory formation from trajectory compression. We clarify these two contributions below.
First, the architectural contribution lies in separately organizing historical context refinement and target SID generation while coupling them through recurrent feedback. Existing backbones either keep historical context fixed during decoding or lack an explicit separation between context refinement and target prediction. LoopGR assigns these computations to distinct modules: the encoder initializes historical memory, the shared decoder refines target SID representations and the lightweight updater uses the intermediate decoder state to refine the memory for subsequent cycles. The conditioning memory itself evolves during recurrent decoding. The updater supports GRU, token mixing, or attention operators. This separation allows history refinement and target generation to use different operators and independently controlled recurrent budgets, so computation can be allocated to each according to its role while prediction feedback keeps the two processes coupled.
Second, the training contribution is to decouple exit readout from trajectory formation and transfer the learned refinement dynamics to a lower-budget model. Jointly optimizing intermediate exit readouts and recurrent trajectories requires the model to make accurate predictions at early cycles, which can cause representations to stabilize too early and limit the gains from additional recurrent steps. RTSD first trains a high-budget teacher with last step supervision, allowing intermediate states to compose a more effective trajectory. It then trains a lower-budget student by aligning predictive distributions and their evolution directions. This separates learning the refinement process from learning to approximate it with fewer recurrent steps. Our additional ablations show that RTSD preserves recommendation quality more effectively than teacher-initialized, task-only fine-tuning (W4/Q2).
(W2) Evaluation beyond Amazon and with longer interaction histories.
We have extended the evaluation to an industrial recommendation dataset with millions of training samples and an item sequence length of 100, corresponding to 400 SID tokens. We train on samples from the first seven days and evaluate predictions on the eighth day. Each historical item is represented by four SID tokens. We compare three baselines, including Lazy Decoder-only, TIGER and BARGE, with four LoopGR variants under this industrial setting. The complete results are reported below.
LoopGR with the Mixing updater achieves the highest Hit@K at all five cutoffs. Relative to TIGER, its improvements are 4.14%, 5.44%, 6.70%, 6.87% and 7.13% at K = 5, 10, 20, 50 and 100, respectively. These results provide additional evidence for LoopGR in industrial recommendation with longer input sequences and across different retrieval depths.