Pith. sign in

REVIEW 4 major objections 4 minor 50 references

Test-Time Optimization of Query Embeddings with Ranking Aware Reward Maximization

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A learned vector lifts dense retrieval by up to 8.36 nDCG@10

desk verdict Solid, useful extension of test-time query-vector distillation with a budget-aware scope analysis; needs error bars and comparisons to TOUR/ReFIT/GQR, but the core claims are plausible and the transfer results hold up. read the letter →

arxiv 2608.12569 v1 pith:4AUCRGN2 submitted 2026-08-12 cs.IR

Tianyu Chen , Jiaxing Wu This is my paper · ORCID
classification cs.IR
keywords test-timeoptimizationdenseretrievalqueryembeddingsknowledgedistillationrankingrewardsembeddingspaceadaptationcatastrophicforgettingrewardbudget
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Test-time ranking scores from a reranker or LLM judge are usually used once and then discarded. This paper proposes TTT-Embed, which distills those scalar scores into a small residual vector added to the frozen query embedding, so the same rewards improve later retrievals. A single scope parameter decides whether the vector is shared by all queries, by one task, or kept private to a query, and the best scope shifts from global to task to query as the reward budget grows. Across five embedding models and fifteen retrieval tasks, the method reports gains up to +8.36 nDCG@10, transfers to unseen queries and tasks, and recovers general capability lost to fine-tuning while leaving model weights untouched. If the claim holds, reranker feedback becomes a reusable test-time state for any embedding system, including closed API models.

What carries the argument

The central object is the scoped residual vector $\boldsymbol{v}_g$ living in the frozen embedding space, with the adapted query $\tilde{\boldsymbol{x}}_q = \operatorname{unit}(\boldsymbol{x}_q + \alpha_g \boldsymbol{v}_g)$. The learning rule is a listwise knowledge-distillation objective in which the teacher is the reward model's softmax over the candidate set and the student is the softmax of embedding similarities; the only free design choice is which rewarded queries share a group $g$. A scope parameter selects global, task, or query-wise sharing, and a parameter-free shrinkage $\alpha_g = n_g/(n_g+1)$, justified as a Bayes estimator, sets how strongly the vector is applied. This object carries the argument because all the information from the reward model is compressed into one additive vector that can be cached and reused against the unchanged ANN index.

What would settle it

Train a task-wise vector on rewarded queries, then evaluate retrieval on a held-out document pool that excludes every candidate that appeared in the training top-$K$ lists (or on documents that never ranked in any rewarded query's top-$K$). If the gain over raw retrieval does not persist on those unseen documents, the claim that the vector reorders the corpus beyond the candidate set fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the scalar relevance scores a reranker assigns to the top $K$ candidates of a dense retriever can be turned into a durable, reusable state: a single residual vector $\boldsymbol{v}_g$ that is added to the query embedding before a fresh search over the unchanged document index. The vector is learned by minimizing a listwise KL divergence between the softmax of the reward scores (teacher) and the softmax of embedding similarities (student) over the retrieved candidates, with ridge regularization. Because normalization is a positive constant per query, the adapted score orders documents exactly like the base score plus the additive term $\alpha_g \boldsymbol{v}_g^\top \boldsymbol{z}_d$; the paper proves this ranking equivalence and uses it to argue that the correction applies corpus-wide, not only to the re-ranked candidate list. The authors report that with a budget of ten rewards per query the shared state lifts nDCG@10 by 6.36 points on average over raw retrieval and beats direct reranking by 3.15 points, and that gains persist on queries and tasks that never supplied rewards. They further claim that a frozen fine-tuned model can recover general capability lost to specialization (up to +8.00 nDCG@10) without any weight update.

Load-bearing premise

The learned vector is optimized to match the reranker on the base retriever's top-$K$ candidates, but at deployment it is used to search the whole corpus; if matching the teacher on those candidates does not also improve the ranking of unseen documents, including documents outside the candidate sets, the reported gains will not transfer.

Editorial extensions

If this is right

  • With shared scopes, a reward budget spent on a subset of queries improves retrieval for all queries in that scope, so the system can spend scarce reranker calls where they help most.
  • The optimal sharing scope depends on budget: global reuse wins at very low budgets, task-wise at intermediate budgets, and query-private vectors at high budgets, giving a concrete deployment rule.
  • Because no weights or index entries change, the method works with closed embedding APIs and with precomputed document indexes, and the learned state is small enough to store.
  • A fine-tuned model can keep its specialization while regaining broad benchmark performance at test time, since the correction is applied outside the model.
  • The additive structure makes the corrected search equivalent to base score plus a per-document bias $\boldsymbol{v}_g^\top \boldsymbol{z}_d$, so retrieved documents are not limited to the original candidate list.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the reported transfer suggests the learned vector captures task-level regularities rather than query memorization; a direct test would be to train on one task family and evaluate on a disjoint domain with no shared vocabulary.
  • Editorial inference: because the method only needs scalar relevance scores and vector outputs, the same residual-state recipe should transfer to other embedding-based search problems such as code search, recommendation, or multimodal retrieval.
  • Editorial inference: the shrinkage rule treats each rewarded query as an independent unit of evidence; if reward quality varies per query, an evidence-weighted or confidence-weighted version of $\alpha_g$ is a natural extension the paper does not explore.
  • Editorial inference: the finding that general-purpose judges produce usable rewards suggests a cheap-judge, high-depth variant could approximate the specialized reranker results at lower API cost.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes TTT-Embed, a method that learns a lightweight residual vector v_g in the frozen output-embedding space of a dense retriever by distilling scalar ranking scores from an external reranker over the retriever's top-K candidate documents. The vector is added to query embeddings at deployment as x_q + alpha_g v_g, with a scope parameter controlling whether the vector is shared globally, per task, or per query, and with a closed-form evidence-adaptive magnitude alpha_g = n_g/(n_g+1). The authors evaluate on 15 MTEB retrieval tasks across five embedding models (three open-weight and two API-only Gemini models), reporting gains up to +8.36 nDCG@10 over raw retrieval, positive transfer to held-out queries and leave-one-task-out settings, and recovery of general MTEB capability after domain-specific fine-tuning, all without weight access or index modification.

Significance. If the results hold, TTT-Embed offers a practical and conceptually clean interface for reusing ranking rewards across queries under a fixed reward budget, with a principled scope trade-off and a parameter-free shrinkage rule. The paper's strengths include a simple and correct ranking-equivalence proposition, a careful nested-budget evaluation protocol that gives every method the same rewarded query-document pairs, evaluation on closed-source embedding APIs, and explicit held-out-query and leave-one-task-out transfer experiments. The main weaknesses are the absence of empirical comparisons to the closest prior test-time query-embedding methods (TOUR, ReFIT, GQR), the lack of any error bars or significance analysis, and the absence of a direct decomposition separating gains from reordering the training candidate set from gains on documents never seen during optimization.

major comments (4)
  1. [§2.2, §4.1, Table 2, deployment paragraph] The objective in Eq. (4) fits v_g only to the teacher distribution over each rewarded query's candidate set C_q, while deployment performs a fresh ANN search over the full corpus. The held-out-query and leave-one-task-out experiments in Table 2 and Figure 5 already provide aggregate evidence that the learned vector helps on queries whose candidate sets were never optimized, but the main fixed-population budget curves do not separate gains from reordering documents inside C_q from gains on documents outside C_q. Please add a direct analysis, such as nDCG@10 computed only over documents not present in any training C_q for rewarded queries, or the overlap between the original top-10 and the adapted top-10. This would substantiate the claim in §4.1 of 'corpus-wide retrieval improvements' and directly address whether the method is primarily a learned reranking bias or a genuinely reusable state for unseen documents.
  2. [Related Work §8, Tables 1 and 2] The paper does not empirically compare against TOUR (Sung et al., 2023), ReFIT (Reddy et al., 2023), or GQR (Uzan et al., 2025), which are the closest prior methods that optimize a test-query embedding from reranker or complementary-retriever scores. In particular, the query-wise variant of TTT-Embed appears to be nearly identical in setting to ReFIT's query-level distillation before a second retrieval pass. Please add matched-budget comparisons to these methods, or explicitly justify why they are not applicable baselines. Without such comparisons, the claimed advantage over existing test-time query-embedding optimization is not established.
  3. [Tables 1, 2, 3 and Figures 2, 5] No error bars, confidence intervals, or significance tests are reported anywhere in the paper. This matters because some of the key margins are small, notably the leave-one-task-out gains of 1.36 for Gemini Embedding 1 and 0.85 for Gemini Embedding 2 in Table 2, and because the query-selection schedule depends on a fixed seed. Please report variance across at least the query-allocation seed, and across sampling seeds for the stochastic LLM judges if they are used in any main result, or state clearly that the main experiments are deterministic and explain why the single fixed seed is representative.
  4. [§4.3, Figure 3] The claim that TTT-Embed 'resolves catastrophic forgetting' is based on a single fine-tuned model (SKILLRET-SFT-0.6B) and a single specialization task. Catastrophic forgetting generally refers to sequential or continual learning, and one experiment is thin support for the phrase 'resolves'. Please soften the claim to 'mitigates' or 'recovers from domain-specialization degradation' and, if possible, add a second specialization scenario or explicitly discuss the scope of the evidence.
minor comments (4)
  1. [§4.3] The phrase 'provides a elegant solution' contains a grammatical error; it should be 'provides an elegant solution.'
  2. [Figure 5 caption] The caption is difficult to read because of the long parenthetical clause and inconsistent punctuation (e.g., 'paired base on shrunken UB' is not defined before that point). Please rewrite the caption to state clearly that the held-out-query curves use a raw-query baseline on the same shrinking evaluation set.
  3. [Abstract and §1] The phrase 'modifications to index' in the abstract should read 'modifications to the index', and the comma after 'This vector is optimized purely from scalar ranking scores' is misplaced; consider rephrasing for clarity.
  4. [§2.1] The notation in Eq. (1) defines B as the total budget and b as the normalized budget, but the text later refers to 'budget b = B/N' and also uses 'budget b' for a reward budget in §5.1. Please introduce 'normalized budget' consistently to avoid confusion with the total budget B.

Circularity Check

0 steps flagged · score 1.0 of 10

No circularity found: the learned vector is distilled from an external reranker, evaluated on held-out labels, and the deployment magnitude is fixed a priori rather than fitted to the target metric.

full rationale

The paper's derivation chain is self-contained with respect to its evaluation labels. The learned vector v_g is optimized purely by KL distillation against a teacher distribution built from an external reranker's scalar scores over the base retriever's top-K candidates (Eqs. 2-4); ground-truth MTEB labels are used only for evaluation, never for selecting rewards or learning v_g. The deployment magnitude alpha_g = n_g/(n_g+1) is justified by a Bayesian normal-means argument (Proposition 1) and is explicitly fixed a priori and never tuned on evaluation labels, so the shrinkage rule is not a fitted input renamed as a prediction. The held-out-query (80/20) and leave-one-task-out experiments train on disjoint queries/tasks, so their transfer gains are not forced by the training objective. The paper's main risk, that matching the teacher distribution over C_q may not guarantee improved ranking for documents outside C_q during the fresh ANN search, is a validity limitation rather than a circular reduction: Proposition 2 only proves ranking equivalence and the paper does not claim that out-of-candidate documents are optimized by construction. No load-bearing self-citations or imported uniqueness theorems appear; references to TOUR, ReFIT, and GQR are related-work attributions for query-private precursors, and the authors do not cite their own prior work to justify the method's central premise. The only arguable blemish is the hand-chosen kappa=1 in the shrinkage prior, which is a modeling choice rather than a circular step, since the paper explicitly states it is never tuned on evaluation labels and evaluates it against oracle-tuned alternatives.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The method introduces no new physical entities; its epistemic burden is carried by six hand-chosen hyperparameters and by strong transfer assumptions about embedding spaces. The most load-bearing is the assumption that top-K candidate distillation generalizes to full-corpus retrieval.

free parameters (6)
  • Teacher temperature tau_T = 0.02
    Sharpens the teacher softmax in Eq (2); fixed across all experiments without sensitivity analysis. With reward scores in [0,1], this makes the teacher nearly one-hot.
  • Student temperature tau_S = 0.05
    Temperature in the student softmax Eq (3); fixed across all settings.
  • Ridge regularization lambda = 0.3
    Weight on ||v||^2 in Eq (4); fixed across models, scopes, and budgets.
  • Adam learning rate = 0.01
    Optimizer setting for Eq (4).
  • Optimization steps = 300
    Full-batch Adam steps used for every learned vector.
  • Prior-to-likelihood noise ratio kappa = 1
    Chosen in Propositions 1 and 3 so that alpha = n/(n+1); a modeling choice, not fitted to data.
assumptions (6)
  • domain assumption Reward model scores are a reliable teacher for ranking quality.
    Eq (2) builds the teacher distribution from reranker or LLM scores; if the reward model is biased, the learned vector inherits the bias. Section 3.3.
  • domain assumption Optimizing the KL objective over the base retriever's top-K candidates transfers to full-corpus ranking.
    Eq (4) sees only C_q, but deployment uses a fresh ANN search over the entire index. The paper does not prove or ablate this transfer. Section 2.2.
  • domain assumption Normal-means model with kappa=1 justifies the alpha = n/(n+1) magnitude rule.
    Proposition 1 is an idealized model, explicitly stated as not assumed when optimizing Eq (4), but the alpha rule is applied in all experiments. Section 2.4 and Appendix 11.
  • domain assumption Task-balanced query weighting matches the evaluation protocol.
    Eq (14) weights tasks equally to prevent large tasks from dominating; this is tied to MTEB's equal task averaging. Section 10.
  • domain assumption The reward budget counts only query-document pairs and ignores inference compute.
    Section 2.1 defines B as the number of scored pairs; the cost of running the reranker and the second ANN search is excluded from budget comparisons.
  • domain assumption Additive perturbations in embedding space can compensate for fine-tuning degradation.
    Section 4.3 assumes that the SKILLRET-SFT model's lost MTEB capability is recoverable by a query-side vector shift.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Test-Time Optimization of Query Embeddings with Ranking Aware Reward Maximization." pith.science (2026). https://pith.science/paper/4AUCRGN2

@misc{pith2026260812569,
  author       = {Pith},
  title        = {Pith review of: Test-Time Optimization of Query Embeddings with Ranking Aware Reward Maximization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4AUCRGN2}},
  note         = {Machine review of arXiv:2608.12569}
}
read the original abstract

Dense retrievers rank documents using vector similarity between a frozen encoder and a precomputed index. While test-time ranking rewards from a reranker or LLM judge can improve results, existing methods discard this signal after a single query. Updating the retriever's weights makes rewards reusable, but this requires parameter access, which is unavailable for closed-source models, and is computationally prohibitive. We propose TTT-Embed (Test-Time Tuning of Embeddings), a framework that distills ranking rewards into a lightweight, learned vector within the output embedding space of a frozen model. This vector is optimized purely from scalar ranking scores assigned to the retriever's own candidate documents, requiring no access to model weights, ground-truth labels, or modifications to index. A single scope parameter controls rewards reuse (global, task, or query), enabling a principled trade-off between reusability and specificity under a fixed reward computation budget. We demonstrate that as the available reward budget scales, the optimal sharing scope shifts dynamically from global-wise to task-wise and finally to query-wise. Evaluated across five embedding models and 15 MTEB retrieval tasks, TTT-Embed improves test-time retrieval by up to +8.36 nDCG@10. Crucially, the learned states generalize effectively to unseen queries (up to +8.57 nDCG@10) and unseen tasks (up to +4.71 nDCG@10). Furthermore, TTT-Embed successfully resolves catastrophic forgetting: by leaving base weights entirely frozen, it recovers degraded general capabilities (up to +8.00 nDCG@10, even surpassing the original base model) while preserving in-domain specialization. These results establish ranking rewards as a reusable test-time state, enabling budget-efficient adaptation for any embedding model, including closed-source APIs.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 15 canonical work pages

  1. [1]

    Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=

    Dense passage retrieval for open-domain question answering , author=. Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=

    2020

  2. [2]

    arXiv preprint arXiv:2007.00808 , year=

    Approximate nearest neighbor negative contrastive learning for dense text retrieval , author=. arXiv preprint arXiv:2007.00808 , year=

    arXiv 2007

  3. [3]

    Sentence-bert: Sentence embeddings using siamese bert-networks , author=. Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) , pages=

    2019

  4. [4]

    arXiv preprint arXiv:1901.04085 , year=

    Passage Re-ranking with BERT , author=. arXiv preprint arXiv:1901.04085 , year=

    arXiv 1901

  5. [5]

    arXiv preprint arXiv:2605.27295 , year=

    Gemini Embedding 2: A Native Multimodal Embedding Model from Gemini , author=. arXiv preprint arXiv:2605.27295 , year=

    arXiv

  6. [6]

    arXiv preprint arXiv:2509.20354 , year=

    Embeddinggemma: Powerful and lightweight text representations , author=. arXiv preprint arXiv:2509.20354 , year=

    arXiv

  7. [7]

    Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=

    Colbert: Efficient and effective passage search via contextualized late interaction over bert , author=. Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=

  8. [8]

    arXiv preprint arXiv:2210.10634 , year=

    Rankt5: Fine-tuning t5 for text ranking with ranking losses , author=. arXiv preprint arXiv:2210.10634 , year=

    arXiv

Show all 50 references
  1. [9]

    Findings of the Association for Computational Linguistics: ACL 2022 , pages=

    ED2LM: Encoder-decoder to language model for faster document re-ranking inference , author=. Findings of the Association for Computational Linguistics: ACL 2022 , pages=

    2022

  2. [10]

    arXiv preprint arXiv:2308.03281 , year=

    Towards general text embeddings with multi-stage contrastive learning , author=. arXiv preprint arXiv:2308.03281 , year=

    arXiv

  3. [11]

    Proceedings of the 47th international ACM SIGIR conference on research and development in information retrieval , pages=

    C-pack: Packed resources for general chinese embeddings , author=. Proceedings of the 47th international ACM SIGIR conference on research and development in information retrieval , pages=

  4. [12]

    arXiv preprint arXiv:2506.05176 , year=

    Qwen3 embedding: Advancing text embedding and reranking through foundation models , author=. arXiv preprint arXiv:2506.05176 , year=

    arXiv

  5. [13]

    arXiv preprint arXiv:2503.07891 , year=

    Gemini embedding: Generalizable embeddings from gemini , author=. arXiv preprint arXiv:2503.07891 , year=

    arXiv

  6. [14]

    Proceedings of the 30th ACM International Conference on Information & Knowledge Management , pages=

    Improving query representations for dense retrieval with pseudo relevance feedback , author=. Proceedings of the 30th ACM International Conference on Information & Knowledge Management , pages=

  7. [15]

    Findings of the Association for Computational Linguistics: ACL 2023 , pages=

    Optimizing test-time query representations for dense retrieval , author=. Findings of the Association for Computational Linguistics: ACL 2023 , pages=

    2023

  8. [16]

    arXiv preprint arXiv:2305.11744 , year=

    ReFIT: Relevance feedback from a reranker during inference , author=. arXiv preprint arXiv:2305.11744 , year=

    arXiv

  9. [17]

    Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    A Large-Scale Study of Reranker Relevance Feedback at Inference , author=. Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  10. [18]

    arXiv preprint arXiv:2510.05038 , year=

    Guided Query Refinement: Multimodal Hybrid Retrieval with Test-Time Optimization , author=. arXiv preprint arXiv:2510.05038 , year=

    arXiv

  11. [19]

    Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval , pages=

    Active query selection for learning rankers , author=. Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval , pages=

  12. [20]

    Machine Intelligence Research , volume=

    A simple yet effective framework for active learning to rank , author=. Machine Intelligence Research , volume=. 2024 , publisher=

    2024

  13. [21]

    Artificial Intelligence and Statistics , pages=

    Online learning to rank with feedback at the top , author=. Artificial Intelligence and Statistics , pages=. 2016 , organization=

    2016

  14. [22]

    Proceedings of the tenth ACM international conference on web search and data mining , pages=

    Unbiased learning-to-rank with biased feedback , author=. Proceedings of the tenth ACM international conference on web search and data mining , pages=

  15. [23]

    arXiv preprint arXiv:2305.05176 , year=

    Frugalgpt: How to use large language models while reducing cost and improving performance , author=. arXiv preprint arXiv:2305.05176 , year=

    arXiv

  16. [24]

    arXiv preprint arXiv:2203.02155 , year=

    Training language models to follow instructions with human feedback , author=. arXiv preprint arXiv:2203.02155 , year=

    arXiv

  17. [25]

    Advances in neural information processing systems , volume=

    Deep reinforcement learning from human preferences , author=. Advances in neural information processing systems , volume=

  18. [26]

    arXiv preprint arXiv:1707.06347 , year=

    Proximal policy optimization algorithms , author=. arXiv preprint arXiv:1707.06347 , year=

    arXiv

  19. [27]

    arXiv preprint arXiv:2402.03300 , year=

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models , author=. arXiv preprint arXiv:2402.03300 , year=

    arXiv

  20. [28]

    arXiv preprint arXiv:2511.13885 , year=

    Taosearchemb: A multi-objective reinforcement learning framework for dense retrieval in taobao search , author=. arXiv preprint arXiv:2511.13885 , year=

  21. [29]

    arXiv preprint arXiv:2501.12948 , year=

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning , author=. arXiv preprint arXiv:2501.12948 , year=

    arXiv

  22. [30]

    International conference on machine learning , pages=

    Test-time training with self-supervision for generalization under distribution shifts , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  23. [31]

    Advances in Neural Information Processing Systems , volume=

    Ttt++: When does self-supervised test-time training fail or thrive? , author=. Advances in Neural Information Processing Systems , volume=

  24. [32]

    arXiv preprint arXiv:2408.03314 , year=

    Scaling llm test-time compute optimally can be more effective than scaling model parameters , author=. arXiv preprint arXiv:2408.03314 , year=

    arXiv

  25. [33]

    arXiv preprint arXiv:1503.02531 , year=

    Distilling the knowledge in a neural network , author=. arXiv preprint arXiv:1503.02531 , year=

    arXiv

  26. [34]

    Journal of the Royal Statistical Society Series C: Applied Statistics , volume=

    The analysis of permutations , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 1975 , publisher=

    1975

  27. [35]

    2012 , publisher=

    Individual choice behavior: A theoretical analysis , author=. 2012 , publisher=

    2012

  28. [36]

    Proceedings of the 24th international conference on Machine learning , pages=

    Learning to rank: from pairwise approach to listwise approach , author=. Proceedings of the 24th international conference on Machine learning , pages=

  29. [37]

    Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Computationally efficient optimization of plackett-luce ranking models for relevance and fairness , author=. Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  30. [38]

    NeurIPS 2023 Foundation Models for Decision Making Workshop , year=

    Policy-gradient training of language models for ranking , author=. NeurIPS 2023 Foundation Models for Decision Making Workshop , year=

    2023

  31. [39]

    Advances in neural information processing systems , volume=

    Differentiable top-k with optimal transport , author=. Advances in neural information processing systems , volume=

  32. [40]

    2004 , publisher=

    Statistics of extremes , author=. 2004 , publisher=

    2004

  33. [41]

    International conference on machine learning , pages=

    Stochastic beams and where to find them: The gumbel-top-k trick for sampling sequences without replacement , author=. International conference on machine learning , pages=. 2019 , organization=

    2019

  34. [42]

    Proceedings of the 2017 conference on empirical methods in natural language processing , pages=

    Task-oriented query reformulation with reinforcement learning , author=. Proceedings of the 2017 conference on empirical methods in natural language processing , pages=

    2017

  35. [43]

    arXiv preprint arXiv:1705.07830 , year=

    Ask the right questions: Active question reformulation with reinforcement learning , author=. arXiv preprint arXiv:1705.07830 , year=

    arXiv

  36. [44]

    Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Reinforcement learning to rank in e-commerce search engine: Formalization, analysis, and application , author=. Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  37. [45]

    SIAM review , volume=

    Optimization methods for large-scale machine learning , author=. SIAM review , volume=. 2018 , publisher=

    2018

  38. [46]

    arXiv preprint arXiv:1412.6980 , year=

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    arXiv

  39. [47]

    arXiv preprint arXiv:2108.08877 , year=

    Sentence-t5: Scalable sentence encoders from pre-trained text-to-text models , author=. arXiv preprint arXiv:2108.08877 , year=

    arXiv

  40. [48]

    arXiv preprint arXiv:2210.07316 , year=

    Mteb: Massive text embedding benchmark , author=. arXiv preprint arXiv:2210.07316 , year=

    arXiv

  41. [49]

    Forty-third International Conference on Machine Learning , year=

    REAL: Regression-Aware Reinforcement Learning for LLM-as-a-Judge , author=. Forty-third International Conference on Machine Learning , year=

  42. [50]

    arXiv preprint arXiv:2605.05726 , year=

    SkillRet: A large-scale benchmark for skill retrieval in LLM agents , author=. arXiv preprint arXiv:2605.05726 , year=

    arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.