Search arXivSearch

arXiv · 2603.22916

GateSID: Adaptive Gating for Balancing Semantic and Collaborative Signals in Recommendation

Abstract

In cold-start scenarios, the scarcity of collaborative signals for new items exacerbates the Matthew effect, undermining platform diversity and posing a persistent challenge in practice. Existing methods augment cold-start items' collaborative signals with semantic information, yet face a collaborative-semantic trade-off: collaborative signals work well for popular items but degrade on cold-start ones, while excessive reliance on semantics ignores collaborative differences. To address this, we propose GateSID, which introduces an adaptive gating network to dynamically balance semantic and collaborative signals based on item maturity. We first discretize multimodal features into hierarchical Semantic IDs (SID) via Residual Quantized VAE, then propose two components: (1) Gating-Fused Shared Attention (GFSA), which fuses attention distributions with gate-regulated weights; (2) Gate-Regulated Contrastive Alignment (GRCA), which enforces stronger alignment for cold-start items while relaxing it for popular ones. Experiments on large-scale industrial datasets demonstrate GateSID's superiority over competitive baselines, with the largest gains on popular items. An online A/B test confirms practical effectiveness: GMV +2.6%, CTR +1.1%, and Order +1.6%, with less than 5ms of additional latency. Beyond the method itself, we conduct a comprehensive exploration of SID in ranking models, systematically studying embedding types, SID configurations, and fusion strategies. We hope this exploration offers some useful insights for the community.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Hai Zhu, Yantao Yu, Lei Shen, Bing Wang, Xiaoyi Zeng. 2026-07-09. GateSID: Adaptive Gating for Balancing Semantic and Collaborative Signals in Recommendation. https://arxiv.org/abs/2603.22916

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks

Foundation models now enable autonomous agents to interact with real-world websites, but existing benchmarks emphasize general-purpose browsing, underrepresent research-oriented environments and scholarly discovery workflows, and often depend on live sites whose changing content and structure undermine reproducibility. arXiv provides a realistic, reproducible, hierarchically structured, information-centric testbed without privacy-sensitive interactions. We introduce WebArxiv, a static-snapshot benchmark comprising 510 time-invariant tasks, each with a unique deterministic ground truth. Its diverse, realistic scholarly tasks go beyond simple information lookup and rule following to emphasize multi-constraint paper retrieval, fine-grained content extraction, and cross-paper comparison. Evaluations of a range of foundation-model-based web agents show that WebArxiv remains challenging. Behavioral analysis reveals that agents over-rely on fixed interaction histories, causing incomplete or repetitive reasoning. We therefore equip agents with a lightweight dynamic-memory mechanism for adaptive retrieval and reasoning over relevant context. The benchmark and code are available at https://anonymous.4open.science/r/74E4423BVNW/README.md.

cs.IR

GreekBarRetrieval: A Benchmark for Greek Statutory Retrieval

Statutory retrieval is necessary for citation-grounded legal question answering, but remains underexplored for Greek. We introduce GreekBarRetrieval, a public retrieval benchmark derived from, and complementing GreekBarBench, which did not include retrieval. The new benchmark comprises 283 bar-exam questions, each accompanied by the facts of the case it refers to, and 6,308 candidate statutory articles to retrieve from. Questions and facts are stated in everyday language, but need to be mapped to the formal terminology of statutes and their abstract legal concepts. A further complication is that not all of the case facts are relevant to each question of a case. Experimenting with three BM25 variants and nine dense retrievers, we find that vanilla dense retrieval far outperforms vanilla sparse retrieval in Recall@100. However, LLM-based query reformulation helps BM25 close that gap, while also improving dense retrieval. With a ten-round ReAct-like LLM reformulation loop that we introduce, BM25 improves further in Recall@100 and obtains the best nDCG and MAP scores of all tested retrievers. Query reformulation also outperforms pseudo-relevance feedback, sparse-dense fusion, and English translation.

cs.IR

Parameterized Dense-Sparse Fusion for Hybrid Retrieval: Tuning a Rank-Score Mix on BEIR SciFact with Qdrant

We study a parameterized hybrid ranker that fuses a dense embedding list and a sparse lexical list. The method has a small, explicit parameter vector: a dense prior $α\in [0,1]$, a score-versus-rank mix $λ\in [0,1]$, an RRF smoothing parameter $κ> 0$, optional list-geometry coefficients that move $α$ per query, and a router margin $τ$ that can turn sparse search off. We grid-search those ranges on SciFact train (809 queries) and freeze the chosen values on SciFact test (300). The tuned rank-score mix ($α= 0.8$, $λ= 0.75$, $κ= 20$) reaches 0.753 nDCG@10 and 0.889 recall@10, outperforming dense BGE (0.742 / 0.871) and equal-weight RRF (0.707 nDCG@10) on that test split. A list-conditioned $α$ adds +0.0006 nDCG; a sparse-off router is rejected by the same train split (any $τ$ that skipped approximately 50% of queries lost nDCG). These coefficients are dataset-specific. Equal RRF with the same models does not beat dense on a nine-zip BEIR macro-average (0.479 vs. 0.519 nDCG@10). Repeating the same train-then-freeze sweep independently on all 20 indexed units beats equal RRF on 20/20 and dense on 16/20 (unit-mean nDCG@10 0.467 vs. 0.462 dense vs. 0.420 RRF). Other corpora should reuse the ranges, not a copy of the SciFact point.

cs.IR