Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Hyperband-based Bayesian Optimization for Black-box Prompt Selection

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A new method, HbBoPs, selects better LLM prompts with fewer API calls by combining a structural-aware Gaussian Process with Hyperband scheduling.

desk verdict A solid, well-tested engineering contribution to black-box prompt selection, with an honest but load-bearing efficiency claim that needs a small fix before the abstract can stand. read the letter →

arxiv 2412.07820 v2 pith:TVU3IKBD submitted 2024-12-10 cs.LG cs.AI

Lennart Schneider , Martin Wistuba , Aaron Klein , Jacek Golebiowski , Giovanni Zappella , Felice Antonio Merra This is my paper · ORCID
classification cs.LGcs.AI
keywords promptselectionblack-boxoptimizationBayesianmulti-fidelityHyperbanddeepkernelGaussianProcesslargelanguagemodelsfew-shotlearning
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

This paper introduces HbBoPs, a method for static black-box prompt selection that jointly chooses instructions and few-shot exemplars from a finite candidate pool. The method pairs a structural-aware deep kernel Gaussian Process with Hyperband as a multi-fidelity scheduler to balance sample-efficiency (fewer prompt evaluations) and query-efficiency (fewer total LLM calls). The central claim, supported by experiments across ten benchmarks and three LLMs, is that HbBoPs outperforms existing full-fidelity and multi-fidelity methods—especially in anytime performance under a limited budget. If true, it means practitioners can identify better-performing prompts for black-box LLMs while spending fewer API calls.

What carries the argument

The central machinery is the structural-aware deep kernel Gaussian Process: a GP whose kernel is a learned feature extractor that processes instruction embeddings and exemplar embeddings through separate feed-forward networks, concatenates them, and projects to a low-dimensional latent space aligned with downstream performance. This surrogate is trained online and used with Expected Improvement to propose candidate prompts within Hyperband brackets, where the number of validation instances serves as the fidelity parameter.

What would settle it

Re-run the HbBoPs experiments with high sampling temperature (e.g., temperature=1.0) or with seed-dependent variability in LLM outputs, and compare the number of true LLM calls needed to reach the same test error as reported; if the cached lower-fidelity evaluations produce inconsistent rankings, HbBoPs's edge over TRIPLE-SH would shrink or vanish.

Watch

Extended reading notes

Core claim

The paper claims that combining a structural-aware deep kernel Gaussian Process (which learns a low-dimensional latent representation from separate embeddings of instructions and few-shot exemplars) with Hyperband (which adaptively allocates validation instances across fidelity levels) yields a prompt selection method that is both sample-efficient and query-efficient. In experiments, HbBoPs achieves the lowest average normalized test error at full budget (0.150) compared to all baselines and competitors, and shows the largest improvements in the low-budget regime, e.g., roughly 35% lower error than the best full-fidelity method (HDBO) and 24% lower than the best multi-fidelity method (TRIPLE-SH) at 0.25 budget. The paper also demonstrates through ablation that each component (deep kernel, structural-awareness, Hyperband, and BO proposal) contributes significantly to the overall performance.

Load-bearing premise

The method's query-efficiency gains assume that LLM outputs are reasonably deterministic, so that cached lower-fidelity evaluations remain valid at higher fidelities; the paper uses temperature=0.5, making this assumption fragile.

Editorial extensions

If this is right

  • HbBoPs can identify well-performing prompts with substantially fewer LLM calls than full-fidelity Bayesian optimization or bandit methods, making black-box prompt selection more practical under API budget constraints.
  • In the low-budget regime (e.g., 0.25 of total calls), the method's gains are largest, suggesting it is especially effective when evaluation budgets are severely limited.
  • The structural-aware deep kernel generalizes across encoder models, indicating the approach is robust to the choice of embedding backbone.
  • The paper's design choices—caching lower-fidelity evaluations, using supersets for higher stages, and selecting the incumbent from full-fidelity evaluations—make Hyperband viable for prompt selection.
  • The method can be extended to evolving candidate pools, enabling integration with mutation-based or iterative prompt optimizers.

Reading between the lines

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

  • The efficiency gains reported are partially contingent on caching lower-fidelity LLM outputs and assuming near-deterministic outputs; with temperature=0.5, stochasticity could undermine this assumption, potentially inflating the reported query-efficiency advantage over methods that do not cache.
  • The paper's focus on exact-match loss on ten benchmarks may not capture settings with more nuanced or open-ended outputs, where the surrogate's learned embeddings may transfer differently.
  • A direct extension would be to test HbBoPs on larger candidate pools (e.g., hundreds or thousands of prompts) to see whether the deep kernel's capacity and Hyperband's bracket structure scale gracefully.
  • The multi-objective extension mentioned in the paper (balancing prompt performance against inference latency) is a natural next step, but is not explored here.
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

3 major / 6 minor

Summary. The paper proposes HbBoPs, a black-box prompt selection method that combines a structural-aware deep kernel Gaussian Process (GP) with Hyperband multi-fidelity scheduling. Prompts are composed of an instruction and a few-shot exemplar; the GP is trained on separate BERT embeddings of the two components and proposes candidates via expected improvement, while Hyperband controls the number of validation instances per evaluation. The authors evaluate HbBoPs against random search, four full-fidelity baselines (vanilla BO, HDBO, BOPCA, EASE, MIPROv2) and two multi-fidelity state-of-the-art methods (TRIPLE-SH, TRIPLE-GSE) on ten tasks and three LLMs, with 30 repeats per scenario, reporting normalized validation/test errors versus total LLM-call budget. They report that HbBoPs achieves lower average errors than all competitors, especially in the anytime regime, and they support this with mixed-effects models, Tukey post-hoc tests, component ablations, and an encoder robustness study.

Significance. The contribution is potentially valuable: to my knowledge it is the first method in static black-box prompt selection to combine a surrogate model with a multi-fidelity scheduler, and the empirical design is unusually thorough (10 tasks x 3 LLMs x 30 runs, with linear mixed-effects models and corrected post-hoc tests, plus ablations of the deep kernel, structural-awareness, and Hyperband design choices). The statistical analysis is a strength: the authors report exact test statistics and p-values rather than only mean curves. However, the headline claim of 'outperforms state-of-the-art methods in both performance and efficiency' is only partially supported: at full budget, the advantage over TRIPLE-SH and TRIPLE-GSE is not statistically significant, and the efficiency claim relies on a caching assumption that is not validated under temperature=0.5. The anytime improvements and component ablations are credible and are the paper's main contribution.

major comments (3)
  1. [Section 5.1; Appendix E.1] The claim that HbBoPs outperforms all methods at full budget is not supported by the paper's own statistical tests. In Appendix E.1, the Tukey post-hoc tests at fraction 1.00 report non-significant differences versus TRIPLE-SH for validation (z=1.87, p=0.236) and test error (z=1.49, p=1.000), and versus TRIPLE-GSE for test error (z=1.34, p=1.000). Table 2 also shows median test improvements over TRIPLE-SH at full budget of -0.006 (Claude 3 Haiku), 0.000 (LLAMA3 8B), and -0.001 (Mistral 7B). The abstract and Section 5.1 should either restrict the 'outperforms' claim to the anytime regime or provide additional evidence (e.g., more repetitions, a different aggregation, or a focused analysis) that HbBoPs also improves final test performance.
  2. [Appendix C; Section 3.3; Section D.2] The query-efficiency advantage is partly built on caching lower-fidelity outputs and reusing them in higher stages. Appendix C justifies this by assuming 'reasonably deterministic outcomes,' but the experimental protocol sets temperature=0.5 for all three LLMs (Section D.2), and Section 2 states that the validation error is observed with noise. Under stochastic sampling, a cached output is one realization; reusing it at higher fidelity biases the paired halving comparisons and understates the number of LLM calls that would be needed to obtain an unbiased estimate at the higher fidelity. The paper should either validate the determinism assumption (e.g., with temperature=0 or fixed seeds) or quantify the sensitivity of the anytime performance curves and the claimed factor-of-eta call reduction to stochasticity, for example by comparing the cached implementation with an implementation that re-queries at each fidelity.
  3. [Section 5.2; Table 2] The text states that 'HbBoPs consistently outperforms TRIPLE-SH in terms of both anytime validation and test error,' but Table 2 shows the opposite at full budget for test error: the median relative improvements are negative or zero for all three LLMs. This overstatement should be corrected, e.g., by limiting the consistency claim to the anytime fractions (0.25 and 0.50) and noting that the full-budget test differences are within noise.
minor comments (6)
  1. [Figure 1 caption] The legend entry 'TRIPLE-SH (Shi et al. 2014)' should read '(Shi et al., 2024)' to match the reference list.
  2. [Appendix C] The passage 'roughly by a factor of eta), In HbBoPs we cache the output...' contains a stray closing parenthesis and a missing sentence break; please revise for clarity.
  3. [Section 4.2] The threshold 'at least four observations' used for training the deep-kernel GP is defined only in the experimental setup; it would be helpful to state it in the method description in Section 3.4 alongside the definition of 'enough' observations.
  4. [Algorithm 1] The expression for n in the for-loop header is typeset with an ambiguous 'l' and 'm'; please use explicit floor/ceiling notation so that the bracket schedule is unambiguous.
  5. [Section 5] The normalization of errors 'by the performance of the worst and best prompt' should specify whether the 'best prompt' is the global optimum over the candidate set P or the best prompt found by any method; otherwise the normalized errors are not uniquely defined.
  6. [Section 6 / Reproducibility] No link to the code is provided; given the complexity of the method and the many design choices, releasing the implementation would substantially aid reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; HbBoPs is a standard surrogate-based multi-fidelity selection method evaluated on held-out test sets.

full rationale

The paper's derivation chain is self-contained and non-circular. The surrogate is a structural-aware deep kernel GP trained on observed prompt validation errors (Eq. 6) and used to propose the next prompt via Expected Improvement (Eq. 7); this is standard BO, and the GP's predictions are not presented as first-principles results but as an acquisition mechanism. Reported test errors come from held-out test sets that are not used during prompt selection, so the headline empirical claim does not reduce to fitted inputs. Hyperband is adopted from Li et al. (2018) as an external, independently established algorithm, and the only self-citation is the BOHB-style combination of BO with Hyperband (Falkner et al., 2018), which is a published method with independent content and is not used to justify the paper's central claims. The stochastic-caching concern in Appendix C is a correctness or validity issue about cost accounting under temperature = 0.5, not a circularity: an optimistic query-efficiency estimate would weaken the efficiency claim but does not make any prediction equivalent to its inputs by construction. No self-definitional quantities, renamed known results, or imported uniqueness theorems were found.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central empirical claim relies on several hand-chosen hyperparameters (bmin, eta, rho, MLP widths, min observations) that are not fitted to benchmark results but are also not systematically tuned; the strongest load-bearing assumption is the validity of caching stochastic LLM outputs across fidelities. No new entities are introduced.

free parameters (5)
  • bmin = 10
    Lower limit on validation instances per prompt evaluation; sets the starting budget in every Hyperband bracket (Section 4.2).
  • eta = 2.0
    Halving parameter of Hyperband, controls elimination rate and number of brackets (Section 4.2).
  • rho = 0.1
    Probability of random interleaving in the BO proposal, taken from BOHB (Appendix D.3).
  • feature_extractor_widths = 64-32-10
    Hidden dimensions of the MLPs in the structural-aware deep kernel (Section 3.2), chosen by hand.
  • min_observations_gp = 4
    Minimum number of observations at a fidelity level before the GP is trained for BO proposals (Section 4.2).
assumptions (6)
  • domain assumption Validation error is a black-box function with homoscedastic Gaussian noise
    Equation (4) and footnote 1; actual noise depends on prompt and is heteroscedastic, especially with stochastic LLM outputs.
  • domain assumption Cached LLM outputs remain valid across fidelity levels
    Appendix C assumes reasonably deterministic outcomes; contradicted by temperature=0.5 in Section D.2.
  • domain assumption Exact match loss on validation set is a sufficient objective for prompt selection
    Equation (1); used for all tasks. Fine for these benchmarks, but not generally.
  • standard math Gaussian process posterior and Expected Improvement formulas
    Section 3.1, Equations (5) and (7); taken from Williams & Rasmussen 2006 and Jones et al. 1998.
  • standard math Hyperband and Successive Halving guarantees from bandit literature
    Section 3.3 and Appendix C; relies on Li et al. 2018, Karnin et al. 2013.
  • ad hoc to paper The structural split into instruction and exemplar embeddings with the chosen MLP helps learning
    Section 3.2; motivated by the hypothesis that distinct components carry different structural information. Supported only by the ablation on these benchmarks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hyperband-based Bayesian Optimization for Black-box Prompt Selection." pith.science (2026). https://pith.science/paper/TVU3IKBD

@misc{pith2026241207820,
  author       = {Pith},
  title        = {Pith review of: Hyperband-based Bayesian Optimization for Black-box Prompt Selection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TVU3IKBD}},
  note         = {Machine review of arXiv:2412.07820}
}
read the original abstract

Optimal prompt selection is crucial for maximizing large language model (LLM) performance on downstream tasks, especially in black-box settings where models are only accessible via APIs. Black-box prompt selection is challenging due to potentially large, combinatorial search spaces, absence of gradient information, and high evaluation cost of prompts on a validation set. We propose HbBoPs, a novel method that combines a structural-aware deep kernel Gaussian Process with Hyperband as a multi-fidelity scheduler to efficiently select prompts. HbBoPs uses embeddings of instructions and few-shot exemplars, treating them as modular components within prompts. This enhances the surrogate model's ability to predict which prompt to evaluate next in a sample-efficient manner. Hyperband improves query-efficiency by adaptively allocating resources across different fidelity levels, reducing the number of validation instances required for evaluating prompts. Extensive experiments across ten diverse benchmarks and three LLMs demonstrate that HbBoPs outperforms state-of-the-art methods in both performance and efficiency.

Figures

Figures reproduced from arXiv: 2412.07820 by the authors.

Figure 1
Figure 1. Normalized error (log scale) of the best prompt per method, averaged over benchmarks. Lower is better. Ribbons represent SE [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Normalized error (log scale) of the best prompt per HbBoPs ablation variant, RS, and vanilla BO, averaged over benchmarks. Lower is better. Ribbons represent SE. ing robustness to the choice of encoder. This is expected, as none were specifically fine-tuned for predicting prompt per￾formance. HbBoPs’s effectiveness stems from its ability to learn a mapping from prompts to performance through the structural-aware DK-… view at source ↗
Figure 3
Figure 3. Visualization of the 768 dimensional BERT [CLS] token embeddings of prompts via a two component t-SNE. Left: Raw, unprocessed features. Middle: Features of a 10 component PCA solution. Right: Latent features (10 dimensional) from the feature extractor of our structural-aware DK-GP. Top row: Train split. Bottom row: Test split. Color indicates the performance of prompts for LLAMA3 8B Instruct on GSM8K. 15 [PITH_FULL… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Scatter plots of the validation and test errors of 250 prompts evaluated with LLAMA3 8B Instruct on GSM8K using differently sized (k = 10, 50, 100, 500) bootstrap samples of validation instances (a) to (d) or the full validation set (e). We can observe that if we use t…
Figure 5
Figure 5. Figure 5: Box plots of the bootstrapped variance estimates of the mean validation error of 250 prompts evaluated with LLAMA3 8B Instruct on GSM8K varying the number of validation instances used to estimate the mean validation error. The empirical results we have presented here f…
Figure 6
Figure 6. Figure 6: Normalized error (log scale) of the best prompt found by each HB incumbent selection mechanism, averaged over benchmarks. Lower is better. Ribbons represent SE. To test whether the choice of selection mechanism does make a difference for the final performance of HB, we…
Figure 7
Figure 7. Figure 7: Normalized error (log scale) of the best prompt found by each HB validation instances sampling variant, averaged over benchmarks. Lower is better. Ribbons represent SE. To test whether the design decisions of using supersets and using the same random validation instanc…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Meta-Prompt Optimization for LLM-Based Sequential Decision Making

    cs.LG 2025-02 conditional novelty 5.0 of 10

    EXPO uses adversarial bandit weighting over LLM-generated prompt variations to optimize the meta-prompt of LLM-based sequential decision-making agents, improving performance on optimization and bandit tasks.

Reference graph

Works this paper leans on

24 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [3]

    Left: Raw, unprocessed features

    Visualization of the 768 dimensional BERT [CLS] token embeddings of prompts via a two component t-SNE. Left: Raw, unprocessed features. Middle: Features of a 10 component PCA solution. Right: Latent features ( 10 dimensional) from the feature extractor of our structural-aware DK-GP. Top row: Train split. Bottom row: Test split. Color indicates the perform...

    work page 2024

  2. [4]

    R., and Levy, O

    Honovich, O., Shaham, U., Bowman, S. R., and Levy, O. Instruction induction: From few examples to natural lan- guage task descriptions. In61st Annual Meeting of the Association for Computational Linguistics, ACL 2023, pp. 1935–1952,

    work page 2023

  3. [5]

    pulling an arm

    Box plots of the bootstrapped variance estimates of the mean validation error of 250 prompts evaluated with LLAMA3 8B Instruct onGSM8Kvarying the number of validation instances used to estimate the mean validation error. The empirical results we have presented here further provide justification for using a multi-fidelity scheduler over the validation inst...

    work page 2024

  4. [8]

    Plum: Prompt learn- ing using metaheuristics

    Pan, R., Xing, S., Diao, S., Sun, W., Liu, X., Shum, K., Zhang, J., Pi, R., and Zhang, T. Plum: Prompt learn- ing using metaheuristics. In Ku, L.-W.and Martins, A. and Srikumar, V . (eds.),Findings of the Association for Computational Linguistics: ACL 2024, pp. 2177–2197,

    work page 2024

  5. [9]

    gradient de- scent

    Pryzant, R., Iter, D., Li, J., Lee, Y ., Zhu, C., and Zeng, M. Automatic prompt optimization with “gradient de- scent” and beam search. In Bouamor, H., Pino, J., and Bali, K. (eds.),Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pp. 7957–7968,

    work page 2023

  6. [10]

    Learning to retrieve prompts for in-context learning

    Rubin, O., Herzig, J., and Berant, J. Learning to retrieve prompts for in-context learning. In Carpuat, M., de Marn- effe, M.-C., and Meza Ruiz, I. V . (eds.),Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), pp. 2655–2671,

    work page 2022

  7. [12]

    Reliable gradient-free and likelihood-free prompt tuning

    Shen, M., Ghosh, S., Sattigeri, P., Das, S., Bu, Y ., and Wor- nell, G. Reliable gradient-free and likelihood-free prompt tuning. In Vlachos, A. and Augenstein, I. (eds.),Findings of the Association for Computational Linguistics: EACL 2023, pp. 2416–2429,

    work page 2023

  8. [13]

    L., Wallace, E., and Singh, S

    Shin, T., Razeghi, Y ., Logan IV , R. L., Wallace, E., and Singh, S. AutoPrompt: Eliciting knowledge from lan- guage models with automatically generated prompts. In Webber, B., Cohn, T., He, Y ., and Liu, Y . (eds.),Proceed- ings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 4222–4235,

    work page 2020

Show all 24 references
  1. [14]

    BBTv2: Towards a gradient-free future with large language models

    Sun, T., He, Z., Qian, H., Zhou, Y ., Huang, X., and Qiu, X. BBTv2: Towards a gradient-free future with large language models. In Goldberg, Y ., Kozareva, Z., and Zhang, Y . (eds.),Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pp. 3916...

    2022

  2. [15]

    GPS: Genetic prompt search for efficient few-shot learning

    Xu, H., Chen, Y ., Du, Y ., Shao, N., Yanggang, W., Li, H., and Yang, Z. GPS: Genetic prompt search for efficient few-shot learning. In Goldberg, Y ., Kozareva, Z., and Zhang, Y . (eds.),Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (EM...

    2022

  3. [16]

    Zhang, T., Wang, X., Zhou, D., Schuurmans, D., and Gon- zalez, J

    URLhttps://arxiv.org/abs/2404.08164. Zhang, T., Wang, X., Zhou, D., Schuurmans, D., and Gon- zalez, J. E. TEMPERA: Test-time prompt editing via reinforcement learning. InThe Eleventh International Conference on Learning Representations,

    arXiv

  4. [20]

    1”, “2”, “3

    Exemplary HB schedule for black-box prompt selection assuming a minimum budget of bmin = 10 validation instances, a maximum number ofn valid = 80validation instances being available in total, and a halving parameter ofη= 2.0. Bracket(s)Stage(i)#Instances(b)#Prompts(n) 3 0 10 8...

    2024

  5. [21]

    forward mode and to select instances for few-shot exemplars. Instructions are generated using APE’s forward mode, where Claude 3 Sonnet (Anthropic, 2024), configured with a temperature of 1.0 and default settings otherwise (as in the main paper for Claude 3 Haiku), produces 10...

    2024

  6. [22]

    We include HDBO (Hvarfner et al.,

    All full-fidelity BO methods are implemented within BoTorch (Balandat et al., 2020). We include HDBO (Hvarfner et al.,

    2020

  7. [23]

    high-dimensional

    to have a simple yet well-performing “high-dimensional” BO baseline. Hvarfner et al. (2024) recently challenged the general belief that vanilla BO does not perform well for high-dimensional functions by training a GP via MAP with priors over kernel and likelihood parameters ad...

    2024

  8. [24]

    (2018) for each proposal with a probability of ρ= 0.1

    Moreover, to hedge against poor model-based proposals, we perform random interleaving as described in Falkner et al. (2018) for each proposal with a probability of ρ= 0.1 . Since a single SH or HB schedule may require less budget than the total pre-defined LLM call budget per ...

    2018

  9. [1978]

    J., Purtell, J., Broman, D., Potts, C., Zaharia, M., and Khattab, O

    11 Hyperband-based Bayesian Optimization for Black-box Prompt Selection Opsahl-Ong, K., Ryan, M. J., Purtell, J., Broman, D., Potts, C., Zaharia, M., and Khattab, O. Optimizing instruc- tions and demonstrations for multi-stage language model programs. In Al-Onaizan, Y ., Bansa...

    2024

  10. [2011]

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-V oss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Lit...

    1901

  11. [2013]

    samples from a probability distribution

    is formally characterized by the following: (1) Losses are i.i.d. samples from a probability distribution. (2) Each arm has a fixed expected lossµ p. (3) The goal is to identify the arm with the lowest expected loss. 17 Hyperband-based Bayesian Optimization for Black-box Promp...

    2016

  12. [2018]

    org/abs/1810.09305

    URL https://arxiv. org/abs/1810.09305. Kushner, H. J. A new method of locating the maximum point of an arbitrary multipeak curve in the presence of noise.Journal of Basic Engineering, 86(1):97–106,

    arXiv

  13. [2021]

    org/abs/2110.14168

    URL https://arxiv. org/abs/2110.14168. Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. BERT: Pre-training of deep bidirectional transformers for lan- guage understanding. In Burstein, J., Doran, C., and Solorio, T. (eds.),Proceedings of the 2019 Conference of the North Am...

    arXiv

  14. [2022]

    and Neville, J

    Schnabel, T. and Neville, J. Symbolic prompt program search: A structure-aware approach to efficient compile- time prompt optimization. In Al-Onaizan, Y ., Bansal, M., and Chen, Y .-N. (eds.),Findings of the Association for Computational Linguistics: EMNLP 2024, pp. 670–686,

    2024

  15. [2023]

    Jones, D

    URL https: //arxiv.org/abs/2310.06825. Jones, D. R., Schonlau, M., and Welch, W. J. Efficient global optimization of expensive black-box functions.Journal of Global Optimization, 13:455–492,

    arXiv

  16. [2024]

    Eriksson, D

    URL https://arxiv.org/abs/2404.02717. Eriksson, D. and Jankowiak, M. High-dimensional Bayesian optimization with sparse axis-aligned subspaces. In de Campos, C. and Maathuis, M. H. (eds.),Proceed- ings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence,...

    arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.