Pith. sign in

REVIEW 3 major objections 6 minor 29 references

The paper shows that keyword search and LLM search solve opposite problems—retargeting versus prospecting—and that a learned gate routing only ~10% of users to the LLM captures most of the benefit at 90% lower cost.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

A gated hybrid retrieval system shows rule-based keywords beat LLM queries for retargeting but lose for prospecting, and routing 10% of users to LLM semantic search raises ad conversions by 27.6%.

T0 review reviewed 2026-08-01 challenge →

load-bearing objection Solid industrial systems paper with a genuine decomposition and gate design, but the headline online lift is a system-level effect, not proof of the gating story. the 3 major comments →

arxiv 2607.23121 v1 pith:ODLFK4HI submitted 2026-07-25 cs.IR cs.CLcs.LG

SMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads

Congfei Zhang , Jingxiao Ma , Xiaodong Liu , Hsiang-wei Chao , Siman Wang , Ge Liu , Shantanu Aggarwal , Vincent Zhang
show 9 more authors
Meghana Missula Rachel Liao Zichu Li Xiao Bai Yunzhi Zhou Yajun Wang Zhe Liu Jinchao Li Yu Zhang
This is my paper
classification cs.IR cs.CLcs.LG
keywords Dynamic Product AdsLLM query generationhybrid retrievalBM25dense ANNadaptive routingquality gateconversion rate
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that in dynamic product ads, rule-generated keyword queries and LLM-generated queries are not interchangeable: keyword queries on a lexical BM25 index win at retargeting (re-surfacing known brands and categories), while LLM queries on a dense vector index win at prospecting (discovering new categories). On that basis the authors build SMART, a hybrid retrieval system that serves every user with the zero-cost keyword pass and uses a lightweight learned gate to audit the keyword results for coverage gaps, sending only about 10% of users to the LLM path. Offline, the gated system matches nearly all of the LLM's semantic-prospecting gains while cutting LLM cost by 90%. In a two-week live A/B test, SMART raised the ad conversion rate by 27.6% over a production embedding-based retriever, with click-through rate flat. The paper's central practical claim is that adaptive, output-auditing routing—not always-on LLM retrieval—is the cost-effective way to get LLM value into ad retrieval.

Core claim

The central discovery is a decomposition plus a routing scheme. Controlled experiments on millions of users show that the two query strategies rank oppositely on the two metrics: keyword+BM25 leads on CatRecall@200 by 15.9–24.1% over LLM-on-BM25, while LLM+ANN leads on Relevance Score by 22.6% over keyword+BM25. Because 73% of users see no metric change when LLM candidates are appended, the benefit is concentrated in a ~10% minority. SMART's quality gate predicts that minority from a coverage-gap feature, gap(u)=1-|cat(Rkw)∩cat(H)|/|cat(H)|, and routes only those users to the LLM path, capturing a Relevance Score of 2.36 (vs 2.39 for always-on LLM+ANN) at 10% of the cost, and lifting online

What carries the argument

The load-bearing mechanism is the learned quality gate, a GBDT classifier over 54 features that audits the output of the cheap keyword pass before deciding whether to invoke the LLM. Its defining signal is the coverage gap: the fraction of a user's historical purchase categories missing from the top-200 BM25 results. The second supporting mechanism is the cross-metric inversion documented in Table 2, which shows the same systems ranked in opposite order by CatRecall@200 (retargeting) and Relevance Score (prospecting); this justifies the explicit assignment of keyword queries to BM25 and LLM queries to ANN. The gate's threshold is tuned to admit ~10% of traffic, the band where CatRecall peaks

Load-bearing premise

The online A/B test attributes the +27.6% conversion lift to the full SMART system, but the treatment differs from the embedding baseline in several components at once—the keyword/BM25 path, the reranker, the diversity layer, and the gated LLM path—so the paper does not isolate the marginal contribution of the gate or the LLM itself.

What would settle it

Run an online A/B test with at least three arms: (1) keyword/BM25 plus reranker only, (2) same plus an ungated random 10% LLM routing, and (3) full SMART with the learned gate. If arm 1 matches the 27.6% conversion lift, or if arm 2 matches arm 3, then the coverage-gap gate is not the source of the gains. A cheaper offline check: compare gate-selected 10% users against a random 10% user sample on Relevance Score; the paper's own numbers (oracle +6.9pp vs GBDT +1.1pp CatRecall@5) already suggest much of the possible lift is left on the table.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Any dynamic-product-ads system with a structured catalog can adopt the template: zero-cost lexical pass for all users, output-auditing gate, gated dense/LLM pass for the ~10% with coverage deficits, and expect most of the semantic-prospecting lift at a fraction of the cost.
  • The retargeting/prospecting inversion gives a diagnostic for when generative query expansion helps or hurts: it helps on dense indices for open-ended intent, and hurts on lexical indices for exact-token retargeting.
  • A 10% routing fraction suffices to reach near-peak quality; routing more users through the LLM actually degrades category recall because noise displaces finite candidate slots (CatRecall falls from 56.6% at 10% routing to 54.5% at 100%).
  • The +27.6% conversion lift with flat CTR implies the hybrid surfaces high-intent discovery items rather than clickbait, so the gains should translate to revenue, not just engagement.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The output-auditing gate is a general pattern beyond ads: any high-throughput retrieval stack with a cheap first stage and an expensive neural second stage can condition the expensive stage on the cheap stage's actual output rather than on input difficulty alone; this is a testable extension in web search and e-commerce.
  • Because the online test swaps multiple components at once (keyword path, reranker, diversity layer, gate), the paper leaves open how much of the 27.6% lift is attributable to the LLM path versus the keyword/BM25 path and reranker; an ablation isolating the gate would be the decisive follow-up.
  • The coverage-gap features derive from a category taxonomy, so the method transfers most directly to domains with structured product metadata; in unstructured domains, one would need a learned surrogate for 'coverage' over the first-stage results.
  • The inverse-U cost curve suggests an operating-point methodology: sweep the routing fraction and pick the knee where marginal quality per additional LLM dollar collapses; production systems can re-tune this continuously.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. SMART proposes a hybrid retrieval architecture for Dynamic Product Ads (DPA) that separates two retrieval objectives: a zero-cost rule-generated keyword path on a BM25 index for retargeting, and a gated LLM-query path on a dense ANN index for semantic prospecting. A GBDT quality gate audits the keyword pass output and routes only about 10% of users to the LLM path. The paper claims, based on offline studies (Tables 1–6, Figure 2) and a 2-week online A/B test (Table 10), that this gated pairing captures most of the semantic-prospecting gains at a 90% reduction in LLM cost, and improves ad conversion rate by +27.6% over a two-tower embedding baseline. The offline evaluation includes a temporal holdout, a delayed-label sensitivity check, and run-to-run stability analysis.

Significance. If the results hold, SMART is a practically significant contribution: it provides production-scale evidence that rule-generated keywords and LLM-generated queries serve complementary retrieval objectives, and that output-auditing gates can be an effective cost-control mechanism for LLM-augmented retrieval. The paper's main strengths are its use of a strict temporal holdout (Table 9), a delayed-label robustness check with bootstrap intervals (Figure 3), run-to-run stability verification (Table 8), and a live-traffic A/B test. The decomposition claim is actionable for any industrial retrieval system with fast heuristic and expensive neural components. However, the load-bearing online attribution is undermined by a multi-component treatment arm, and the offline gating evaluation contains a fit-then-evaluate loop on the same metric used for training and threshold selection.

major comments (3)
  1. [§5.4, Table 10] The online A/B test compares full SMART against a two-tower U2I ANN baseline, but the treatment differs from the baseline in at least four components at once: the rule-generated keyword/BM25 path (Algorithm 1), the LambdaMART reranker, the MMR diversity layer, and the GBDT gate plus LLM/ANN path. The abstract and conclusion attribute the +27.6% conversion-rate lift to adaptive LLM routing and the associated 90% cost reduction, but no online arm isolates the gate or even the LLM path. Tables 1–2 show that the keyword/BM25 path alone is strong on retargeting, so a substantial portion of the lift could plausibly come from components independent of the LLM gate. The paper should either add an online ablation (e.g., keyword+BM25+reranker without the LLM path, or a fixed/random-gate 10% arm) or reframe the claim as a full-system effect rather than evidence for the adaptive gating design.
  2. [§4.2, §5.1, Table 6] The routing gate is trained on labels defined as 'SMART beats keywords on CatRecall@200' across two generation cycles, and the same CatRecall@200 metric is then used both to sweep the gate threshold in Figure 2 and to report the Gated-10% row in Table 6. This is a fit-then-evaluate loop on the same quantity: the threshold is selected to maximize the very metric on which the gate was trained, which can inflate the reported +1.1% CatRecall@200 gain and the associated Relevance Score trajectory. Table 9's temporal validation is encouraging but only reports AUC and CatRecall@5 lift; it does not validate the headline CatRecall@200 and Relevance Score at the 10% operating point on held-out days. Please report held-out-day gated performance (with confidence intervals) for the same metrics used to select the operating point.
  3. [§3.1, Tables 1–2, Table 6] The core decomposition result (Table 2) and the gated operating point (Table 6) are presented as point estimates without uncertainty. The cross-metric inversion — CatRecall@200 and Relevance Score ranking the same systems in opposite directions — is load-bearing for the architecture, yet no bootstrap intervals, permutation tests, or significance statements accompany these orderings. Furthermore, Relevance Score is an LLM-as-a-judge metric using Gemini 3 Flash, the same model family that powers the generative path (Gemini 2.5 Flash Lite). The paper states the outputs are anchored against real conversions, but no such anchoring evidence is shown for the offline tables. Please provide intervals for the relevant orderings and an independent judge (e.g., human-rated relevancy or a conversion-anchored proxy) for the prospecting metric.
minor comments (6)
  1. [§4.2, Eq. (2)] Equation (2) introduces Rθ(·) but this notation is not defined in the text; Algorithm 2 uses R_kw and R_smart inconsistently with Eq. (1). Please align notation.
  2. [Table 5] The feature names in Table 5 (e.g., 'Product set density', 'Result diversity') are not formally defined. A one-sentence definition for each would make the gate's behavior reproducible.
  3. [Figure 3] The x-axis label 'T +1 +2 +3 +4 +5 +6 +7' is visually ambiguous; it should read 'Label window: purchases on days T..T+k' with explicit tick values.
  4. [Abstract, §3.1] The abstract claims 'controlled experiments on millions of users', but the offline diagnostic in §3.1 reports 20K unique users. Clarify whether 'millions' refers to the online A/B or a different population, to avoid an apparent inconsistency.
  5. [Table 10] The online A/B table reports CR and CTR relative only, with no absolute values, user counts, or confidence intervals. Reporting at least the sample size and a bootstrap/credible interval for the CR lift would materially strengthen the production claim.
  6. [§6.2] The limitations section is honest about temporal horizon and label censoring, but should explicitly acknowledge that the online A/B does not decompose the effect by component; adding this caveat would align the conclusions with the evidence.

Circularity Check

1 steps flagged

Mild same-metric loop in offline gate evaluation; online +27.6% CR result independent and breaks full circularity

specific steps
  1. fitted input called prediction [§4.2 Adaptive Quality Gate Mechanism; §5.1 Operational Quality-Cost Threshold Sweeps; Table 6]
    "Training labels are mapped as positive via an OR-of-ORs construction if SMART beats keywords on CatRecall@200 across either of two independent cycles ... CatRecall@200 demonstrates an inverse-U curve, peaking at 56.6% within the 10% routing band ... Setting the gate operating threshold to route exactly 10% of incoming traffic captures the peak CatRecall@200."

    The gate's training label is exactly whether SMART beats keywords on CatRecall@200, i.e., a thresholded version of Eq. (1) with m = CatRecall@200. The operating threshold is then selected from a sweep over the same CatRecall@200 curve (Figure 2), and Table 6 reports CatRecall@200 at that selected 10% point as the offline result. Thus the reported +1.1% CatRecall@200 for Gated 10% is a selected value of the objective used to fit the gate and pick θ, not an independent confirmation that gating preserves retargeting quality. The held-out-day evaluation (Table 9, AUC/CatRecall@5) and the online conversion A/B (Table 10, +27.6% CR) use different target quantities and therefore break full circularity, limiting this step to the offline CatRecall@200 operating-point claim.

full rationale

The paper's core decomposition (keyword+BM25 vs LLM+ANN) is an empirical result from controlled offline comparisons and is not definitionally forced: CatRecall@200 and Relevance Score are distinct metrics and rank systems in opposite directions (Table 2). The online +27.6% conversion-rate claim comes from a 2-week A/B against a production U2I embedding baseline, a target (attributed conversions) that is not the same quantity as the gate's CatRecall@200 training label; hence the central production claim does not reduce to the training objective. The only load-bearing same-quantity loop is the offline gate evaluation: labels are defined as 'SMART beats keywords on CatRecall@200', the 10% threshold is chosen from the CatRecall@200 sweep, and Table 6 reports CatRecall@200 at that point. This is a fit/select-then-evaluate loop on one metric, though the paper does provide a held-out-day temporal test (Table 9) and an independent online conversion test (Table 10). No load-bearing self-citations, imported uniqueness theorems, or ansatz-via-citation patterns were found; references to TIGER, SigLIP2, and Faiss are external. The paper itself flags the remaining attribution gap in §6.2 ('attribution of online conversions seeded by each retrieval path remain future work'), which is an experimental-validity limitation rather than circularity. Proportionate score: 3, because the offline CatRecall@200 result is partially selected on its own objective, but the central online claim has independent content.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

SMART introduces no new physical or theoretical entities; its load-bearing assumptions are measurement choices (LLM judge, category-recall proxy, constructed gate labels) and the cleanliness of the production A/B comparison.

free parameters (5)
  • gate_threshold_theta (routing fraction) = 10% of users
    Selected from the quality-cost sweep (Figure 2, Table 6) based on peak CatRecall@200 and Relevance Score gains; the operating point is fit to the evaluation data.
  • engagement_weights (purchase/add-to-cart/view) = 3/2/1
    Hand-chosen weights in Algorithm 1 and the reranker supervision label (§4.1, §4.3); no sensitivity analysis is given.
  • llm_query_count = 3
    Single structured LLM API call outputs exactly three prospecting queries (§4.3); the choice of three is not varied.
  • keyword/LLM candidate slot split = 170 keyword / 30 LLM slots
    Used in the hybrid offline configuration and qualitative cases (§5.2); no ablation justifies 170/30.
  • keyword_query_budget_n = not specified
    Algorithm 1 takes a budget n; the deployed value and its effect are not reported.
axioms (4)
  • domain assumption LLM-as-a-judge (Gemini 3 Flash) 0–3 Relevance Scores are a valid measure of prospecting quality.
    Defined in §3.1 and used for the headline prospecting gains (Table 2, Figure 2); §6.2 admits no human validation or multi-model consensus was used.
  • domain assumption CatRecall@200 on Google Product Categories is a valid proxy for retargeting success.
    Chosen in §3.1 because exact product-ID recall is sparse; assumes category-level recovery drives conversions.
  • ad hoc to paper The OR-of-ORs training label ('SMART beats keywords on CatRecall@200 across two independent cycles') captures stable user-level benefit from the LLM path.
    Constructed in §4.2; if the two cycles are correlated or noisy, the gate optimizes a proxy that may diverge from true incremental conversion value.
  • domain assumption The 2-week A/B test was uniformly randomly split and attributed conversions are unbiased.
    Asserted in §5.4; no confidence intervals, pre-registration, or bucket diagnostics are reported.

reviewed 2026-08-01 · how reviews work

0 comments
Cite this review

Pith. "Pith review of SMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads." pith.science (2026). https://pith.science/paper/ODLFK4HI

@misc{pith2026260723121,
  author       = {Pith},
  title        = {Pith review of: SMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ODLFK4HI}},
  note         = {Machine review of arXiv:2607.23121}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Dynamic Product Ads (DPA) require retrieving relevant items from multi-million product catalogs, balancing two competing objectives: retargeting (re-surfacing known interests) and prospecting (discovering new categories). While Large Language Models (LLMs) capture semantic intent better than traditional embedding models, deploying them at scale introduces prohibitive inference costs and lexical mismatch issues. Through controlled experiments on millions of users, we demonstrate a critical retrieval decomposition: rule-generated queries excel at retargeting on a lexical BM25 index, while LLM-generated queries excel at prospecting on a dense ANN index. Building on this, we propose SMART (SeMantic-aware Adaptive ReTrieval). To manage costs, a lightweight quality gate identifies coverage gaps in initial keyword results, adaptively routing only the ~10% of users who benefit from semantic prospecting to the LLM path. Offline evaluation demonstrates that this gated approach captures the bulk of semantic prospecting gains in Relevance Score while maintaining competitive re-targeting performance at a 90% reduction in LLM costs. Finally, in a 2-week online A/B test at Snap, SMART improved the ad conversion rate by +27.6% over a strong embedding-based baseline.

Figures

Figures reproduced from arXiv: 2607.23121 by Congfei Zhang, Ge Liu, Hsiang-wei Chao, Jinchao Li, Jingxiao Ma, Meghana Missula, Rachel Liao, Shantanu Aggarwal, Siman Wang, Vincent Zhang, Xiao Bai, XiaoDong Liu, Yajun Wang, Yunzhi Zhou, Yu Zhang, Zhe Liu, Zichu Li.

Figure 1
Figure 1. Figure 1: SMART system architecture. All users receive rule-generated queries for BM25 retargeting. The quality gate evaluates [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Quality–cost trade-off across LLM routing fraction. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Sensitivity of CatRecall@200 to the purchase-label [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

29 extracted references · 8 linked inside Pith

  1. [1]

    Kenya Abe, Kunihiro Takeoka, Makoto P Kato, and Masafumi Oyamada. 2025. LLM-based Query Expansion Fails for Unfamiliar and Ambiguous Queries. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. 3035–3039

    2025

  2. [2]

    Eliran Abutbul, Yohay Kaplan, Naama Krasne, Oren Somekh, Or David, Omer Duvdevany, and Evgeny Segal. 2023. Audience Prospecting for Dynamic-Product- Ads in Native Advertising. In2023 IEEE International Conference on Big Data (BigData). IEEE, 1571–1580

    2023

  3. [3]

    Pranjal Aggarwal, Aman Madaan, Ankit Anand, Srividya Pranavi Potharaju, Swaroop Mishra, Pei Zhou, Aditya Gupta, Dheeraj Rajagopal, Karthik Kappagan- thu, Yiming Yang, et al. 2024. AutoMix: Automatically Mixing Language Models. Advances in Neural Information Processing Systems37 (2024), 131000–131034

    2024

  4. [4]

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avi Sil, and Hannaneh Hajishirzi. 2024. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. InInternational conference on learning representations, Vol. 2024. 9112–9141

    2024

  5. [5]

    Luiz Bonifácio, Hugo Abonizio, Marzieh Fadaee, and Rodrigo Nogueira. 2022. InPars: Data Augmentation for Information Retrieval using Large Language Models.arXiv preprint arXiv:2202.05144(2022)

    Pith/arXiv arXiv 2022

  6. [6]

    Christopher JC Burges. 2010. From RankNet to LambdaRank to LambdaMART: An Overview.Learning11, 23-581 (2010), 81. SMART Retrieval

    2010

  7. [7]

    Jaime Carbonell and Jade Goldstein. 1998. The Use of MMR, Diversity-Based Reranking for Reordering Documents and Producing Summaries. InProceedings of the 21st Annual International ACM SIGIR Conference. 335–336

    1998

  8. [8]

    Lingjiao Chen, Matei Zaharia, and James Zou. 2023. FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance.arXiv preprint arXiv:2305.05176(2023)

    Pith/arXiv arXiv 2023

  9. [9]

    Mengxiang Chen, Zhouwei Zhai, and Jin Li. 2026. Probe-then-Plan: Environment-Aware Planning for Industrial E-commerce Search.arXiv preprint arXiv:2603.15262(2026)

    Pith/arXiv arXiv 2026

  10. [10]

    Gordon V Cormack, Charles L A Clarke, and Stefan Buettcher. 2009. Reciprocal rank fusion outperforms condorcet and individual rank learning methods. In Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 758–759

    2009

  11. [11]

    Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep Neural Networks for YouTube Recommendations. InProceedings of the 10th ACM conference on recommender systems. 191–198

    2016

  12. [12]

    Dujian Ding, Ankur Mallick, Chi Wang, Robert Sim, Subhabrata Mukherjee, Victor Rühle, Laks Lakshmanan, and Ahmed H Awadallah. 2024. Hybrid LLM: Cost-Efficient and Quality-Aware Query Routing. InInternational Conference on Learning Representations, Vol. 2024. 41348–41366

    2024

  13. [13]

    Xin Dong, Sen Jia, Ming Rui Wang, Yan Li, Zhenheng Yang, Bingfeng Deng, and Hongyu Xiong. 2025. COEF-VQ: Cost-Efficient Video Quality Understanding through a Cascaded Multimodal LLM Framework. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 4387–4395

    2025

  14. [14]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2025. The Faiss Library.IEEE Transactions on Big Data(2025)

    2025

  15. [15]

    Luyu Gao, Xueguang Ma, Jimmy Lin, and Jamie Callan. 2023. Precise Zero-Shot Dense Retrieval without Relevance Labels. InProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 1762–1777

    2023

  16. [16]

    Gemini Team, Google. 2023. Gemini: A Family of Highly Capable Multimodal Models.arXiv preprint arXiv:2312.11805(2023)

    Pith/arXiv arXiv 2023

  17. [17]

    Nan Hu, Han Li, Jimeng Sun, Lu Wang, Fangkai Yang, Bo Qiao, Pu Zhao, David Dai, Mengyu Liu, Yuefeng Zhan, et al. 2026. AdNanny: One Reasoning LLM for All Offline Ads Recommendation Tasks.arXiv preprint arXiv:2602.01563(2026)

    arXiv 2026

  18. [18]

    Rolf Jagerman, Honglei Zhuang, Zhen Qin, Xuanhui Wang, and Michael Bender- sky. 2023. Query expansion by prompting large language models.arXiv preprint arXiv:2305.03653(2023)

    Pith/arXiv arXiv 2023

  19. [19]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. LightGBM: A Highly Efficient Gradient Boosting Decision Tree.Advances in neural information processing systems30 (2017)

    2017

  20. [20]

    Yi Luan, Jacob Eisenstein, Kristina Toutanova, and Michael Collins. 2021. Sparse, Dense, and Attentional Representations for Text Retrieval.Transactions of the Association for Computational Linguistics9 (2021), 329–345

    2021

  21. [21]

    Gonzalez, M

    Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E. Gonzalez, M. Waleed Kadous, and Ion Stoica. 2024. RouteLLM: Learning to Route LLMs with Preference Data.arXiv preprint arXiv:2406.18665(2024)

    Pith/arXiv arXiv 2024

  22. [22]

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan H Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Q Tran, Jonah Samost, Maciej Kula, Ed H Chi, and Maheswaran Sathiamoorthy. 2023. Recommender Systems with Generative Retrieval. InAdvances in Neural Information Processing Systems (NeurIPS)

    2023

  23. [23]

    2009.The Probabilistic Relevance Frame- work: BM25 and Beyond

    Stephen Robertson and Hugo Zaragoza. 2009.The Probabilistic Relevance Frame- work: BM25 and Beyond. Vol. 4. Now Publishers Inc

    2009

  24. [24]

    Tantivy Contributors. 2023. Tantivy: A Full-Text Search Engine Library Written in Rust. https://github.com/quickwit-oss/tantivy

    2023

  25. [25]

    Michael Tschannen, Alexey Gritsenko, Xiao Wang, Muhammad Ferjad Naeem, Ibrahim Alabdulmohsin, Nikhil Parthasarathy, Talfan Evans, Lucas Beyer, Ye Xia, Basil Mustafa, et al. 2025. SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features.arXiv preprint arXiv:2502.14786(2025)

    Pith/arXiv arXiv 2025

  26. [26]

    Han Wang, Mukuntha Narayanan Sundararaman, Onur Gungor, Yu Xu, Krishna Kamath, Rakesh Chalasani, Kurchi Subhra Hazra, and Jinfeng Rao. 2024. Improv- ing Pinterest Search Relevance Using Large Language Models.arXiv preprint arXiv:2410.17152(2024)

    Pith/arXiv arXiv 2024

  27. [27]

    Liang Wang, Nan Yang, and Furu Wei. 2023. Query2doc: Query Expansion with Large Language Models. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 9414–9423

    2023

  28. [28]

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. 2023. Judg- ing LLM-as-a-Judge with MT-Bench and Chatbot Arena.Advances in neural information processing systems36 (2023), 46595–46623

    2023

  29. [29]

    Yutao Zhu, Huaying Yuan, Shuting Wang, Jiongnan Liu, Wenhan Liu, Chenlong Deng, Haonan Chen, Zheng Liu, Zhicheng Dou, and Ji-Rong Wen. 2025. Large Language Models for Information Retrieval: A Survey.ACM Transactions on Information Systems44, 1 (2025), 1–54

    2025

This paper was first reviewed by deepseek-v4-flash on August 1, 2026.