Pith. sign in

REVIEW 3 major objections 4 minor 144 references

Weak supervision on abundant unlabeled contexts beats label-only training for downstream decisions, provided the weak and strong feature representations share little overlap.

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 ↗

T0 review · deepseek-v4-flash

2026-08-01 15:18 UTC pith:E3ONJNHA

load-bearing objection The W2S upper bound is a real contribution, but the outperformance certificate leans on a calibration assumption that fails for the paper's own experiments — the framework is sound, the sufficient conditions are narrower than claimed. the 3 major comments →

arxiv 2607.18467 v1 pith:E3ONJNHA submitted 2026-07-20 cs.LG

Weak-to-Strong Learning in Decision Making

Jingwei Ji , Renyuan Xu This is my paper
classification cs.LG
keywords weak-to-strong learningcontextual stochastic optimizationdecision-aware trainingcorrelation dimensionlabel scarcityunlabeled dataplug-in policyexcess decision risk
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.

This paper asks whether weak supervision can improve decision making, not just prediction. The setting is contextual stochastic optimization: choose an action for each context to minimize expected cost, where outcome labels are scarce, unlabeled contexts are abundant, and two pre-trained feature representations (a weak and a strong one) are available. The paper proposes training a weak model on the labeled data, using it to produce soft outcome distributions on the unlabeled contexts, and training the strong model on those pseudo-distributions. Its main theorem gives a non-asymptotic upper bound on the induced policy's excess decision risk; comparing that bound with a lower bound for directly training the strong model on labels alone yields an explicit certificate (Corollary 1) for when the weak-to-strong policy is better. The governing quantity is the correlation dimension between the weak and strong features: the smaller the overlap, the more the abundant unlabeled data wash out teacher errors instead of transmitting them to the student.

Core claim

W2S training — fit a weak model on scarce labels, generate soft outcome distributions on abundant unlabeled contexts, train a strong model on them — can achieve lower expected excess decision risk than training the strong model on labels alone. Corollary 1 certifies this when imitation error, approximation errors, and overlap cost C_t d_{s∧w}/n are small enough to leave a positive margin M_n and N is large enough. The mechanism: weak-teacher error propagates mainly along the d_{s∧w} shared feature directions, while the non-overlap part decays as d_s(d_w−d_{s∧w})/N, so abundant unlabeled contexts dilute, rather than transmit, teacher errors.

What carries the argument

The correlation dimension d_{s∧w} — the squared Frobenius norm of the whitened cross-covariance between the weak and strong feature maps, a number between 0 and min{d_s, d_w} measuring how many aligned directions the two representations share — is the quantity that decides whether W2S beats strong-only training. The proof decomposes the W2S excess decision risk into four terms: imitation error (how well the strong class reproduces the weak teacher's logits), an N^{-1/2} statistical error from the unlabeled sample, a teacher-error term proportional to E‖Θ̂_w−Θ*_w‖²_F times (d_{s∧w}+d_s(d_w−d_{s∧w})/N), and approximation errors ρ_s+ρ_w. Only the overlap directions carry persistent teacher erro

Load-bearing premise

The comparison relies on the assumption that for every context, the distance between the decisions made by the estimated and the true strong model is at least a fixed positive multiple (γ) of the distance between their predicted outcome distributions; the paper assumes this inequality rather than deriving or testing it, and the strong-only lower bound — the anchor for the W2S-vs-strong comparison — collapses if it fails.

What would settle it

In the synthetic newsvendor with controlled overlap r, the theory predicts OPR strictly decreases as r increases (at fixed n,N) and OPR>1 whenever the certified margin M_n>0 holds with N above the threshold in (57); measuring these two relations across trials and finding a violation would falsify the claim.

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

If this is right

  • In label-scarce operational settings (inventory, pricing, moderation routing), a small weak model can supervise a large strong model across unlabeled contexts and deliver a decision policy with no higher expected cost than fine-tuning the strong model on labels alone.
  • The certified regime is self-limiting: gains shrink as labeled samples grow (the margin M_n shrinks) and as the weak–strong overlap dimension d_{s∧w} grows; practitioners can therefore predict when W2S stops paying off.
  • Given a positive margin, the unlabeled budget that certifies W2S scales roughly as max{d_s, n²} up to constants — an explicit data requirement that can be forecast before training.
  • Exact realizability rules out zero overlap (Appendix B.11): d_{s∧w} is bounded below by a ratio of parameter norms, so unlabeled data alone cannot make W2S arbitrarily accurate.
  • The framework is agnostic to how the weak teacher is trained: any procedure with a mean-squared parameter error bound (MLE is one example) can serve as the teacher.

Where Pith is reading between the lines

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

  • The strong-only lower bound in Corollary 1 leans on a calibration inequality (47) that the paper neither derives nor tests; if that inequality fails in an application, the certificate loses its anchor while the W2S upper bound itself may still hold — practitioners should check it before relying on the comparison.
  • Because d_{s∧w} is computable from feature covariances before any training, it could serve as a model-selection diagnostic for choosing which weak/strong pair to use; the paper treats it as an analysis parameter rather than a selection tool.
  • The dilution mechanism suggests a transferable intuition for sequential decisions: if teacher errors behave like noise along non-overlapping state-feature directions, reinforcement learning may inherit teacher bias only along overlap directions — a testable extension the paper leaves open.
  • Both experiments train on smoothed surrogate decision rules but evaluate on the exact rule; an enterprising reader could check whether the OPR>1 regime survives exact non-smooth optimization, to separate the theory's effect from the surrogate's.

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 / 4 minor

Summary. The paper proposes a weak-to-strong (W2S) training framework for contextual stochastic optimization. A weak model is trained on scarce labeled data and used to generate soft outcome distributions on many unlabeled contexts; a strong model is then trained on these pseudo-distributions and evaluated by the downstream decision risk of the induced plug-in policy. The main technical contribution is Theorem 1, a non-asymptotic upper bound on the W2S excess risk decomposed into imitation error, unlabeled-sample statistical error, weak-teacher estimation error, and approximation error, with the teacher error governed by a newly defined correlation dimension d_{s∧w} between weak and strong feature representations. Theorem 2 gives a lower bound for the strong-only benchmark that trains directly on labeled data. Corollary 1 combines the two bounds into sufficient conditions, stated up to universal constants, under which W2S is certified to outperform strong-only training. The paper also reports a synthetic newsvendor experiment and a comment-moderation experiment on real text data, finding qualitative agreement with the predicted regimes: W2S helps most when labeled data are scarce, unlabeled data are abundant, and feature overlap is small.

Significance. If the results hold, this would be a useful first theoretical treatment of weak-to-strong transfer in decision-making rather than prediction, with an appealing interpretable mechanism: the overlap dimension d_{s∧w} determines how much teacher error persists. The paper has notable strengths: the risk decomposition is careful, supporting lemmas are provided in the appendix, and the no-free-lunch bound (Lemma 12) shows the authors explicitly guard against the misleading conclusion that d_{s∧w}=0 makes teacher error vanish entirely. The synthetic experiments control the overlap dimension directly, and the comment-moderation study tests the qualitative predictions in a realistic setting. However, the central W2S-versus-strong certificate depends critically on an unverified and, for natural cost functions, false calibration assumption in Theorem 2, and the comparison is made only up to unspecified universal constants. Thus the significance is currently conditional: the mechanism is plausible and the proof structure is credible, but the benchmark lower bound needs substantial repair or restriction before the paper's main claim is established.

major comments (3)
  1. [§4.1, Theorem 2, Eq. (47); used in Corollary 1, Eq. (56)] The calibration inequality (47) is the load-bearing lower anchor of the entire comparison, but it is not implied by Assumptions 1–2 and is false for a standard class of costs. For quadratic cost c(w,ξ)=(w−ξ)², the plug-in decision is the conditional mean w(P)=E_P[ξ]; for K≥3 there are distinct distributions on the same support with equal means, so the left side of (47) is zero while the right side is positive, forcing γ=0. The same structural failure occurs for any finite action set, including the comment-moderation experiment in §5.2, where w(P) is piecewise constant and hence has a nontrivial nullspace. Without a positive γ, Theorem 2's lower bound (50) degenerates to only the approximation term ρ_s and the 1/n statistical term that anchors Corollary 1 disappears. The paper needs either a lower-bound proof that does not require injectivity of the decision map P↦w(P), or an explicit res
  2. [§4.3, Eq. (53)–(58)] The comparison in Corollary 1 is made 'up to universal constants,' but the margin condition (56) and the subsequent statements treat the comparison as an exact inequality. Since the constants in the upper and lower bounds are not tracked, a positive residual margin M_n>0 computed from the displayed expressions does not logically imply (58) unless the hidden constants happen to satisfy the right inequality. As written, the result is a rate-level comparison, not a certificate of outperformance for fixed problem instances. The paper should either track all constants in the sufficient condition or state the conclusion as an asymptotic/rate result rather than a finite-sample guarantee.
  3. [§5.1–§5.2] The experiments are presented as validating the theoretical conditions of Corollary 1, but neither experiment satisfies the assumptions needed for the main certificate. The synthetic newsvendor uses a non-smooth, non-strongly-convex cost, which violates Assumption 2, and the comment-moderation experiment explicitly states that 'this experiment does not enforce any of the assumptions used in the theory.' In particular, the calibration inequality (47) is never checked in either experiment. The empirical results may still be useful as heuristic evidence, but they cannot be cited as confirmation of the sufficiency conditions in Corollary 1. The paper should either design an experiment that satisfies all assumptions of the theorem or clearly restrict the claim to qualitative consistency.
minor comments (4)
  1. [Corollary 1, Eq. (57) and following paragraph] The text says (57) 'scales qualitatively as N≳max{ds, n²}.' This is not generally implied by the displayed formula. If the approximation errors ρ_s, E_imit, and E_approx are strictly positive so that M_n has a positive limit as n→∞, the required N from the first term is O(1), not O(n²); the n² scaling only emerges in a special regime where the benchmark 1/n term is the only positive contribution to M_n. Please clarify the regime in which the n² statement is intended.
  2. [Definition 1, Eq. (2)] The quantity d_{s∧w} is called a 'dimension' but it is a squared Frobenius norm and is generally not an integer. This is potentially confusing; consider renaming it 'overlap' or 'effective correlation' or explicitly noting that it is a continuous measure, not a rank.
  3. [§2.2, Eq. (11)] There is a typo in 'o utperforming ratio' in the sentence defining OPR. Otherwise the OPR definition is clear.
  4. [Theorem 2, item 1] Inequality (47) uses norms of two objects of different types: a decision vector and a probability vector. Please specify explicitly that the same Euclidean norm is meant in both cases, or use separate notation if different norms are intended.

Circularity Check

0 steps flagged

No significant circularity: the W2S upper bound, strong-only lower bound, and comparison are derived from stated assumptions rather than fitted to the conclusion.

full rationale

The paper's derivation is self-contained in the relevant sense. Definition 1 defines the correlation dimension d_{s∧w} from population feature covariances, not from the excess-risk quantities being predicted, and this quantity enters Theorem 1 through explicit whitened-matrix calculations (Lemma 5) rather than being tuned to force the result. The W2S upper bound (Theorem 1) is obtained by decomposing the risk into imitation error, unlabeled-sample statistical error, teacher-error propagation, and approximation error; each term is bounded using the stated Assumptions 1–4. The no-free-lunch Lemma 12 explicitly prevents d_{s∧w} from being made arbitrarily small under exact realizability, so the overlap dimension is not an unconstrained knob. The strong-only benchmark lower bound (Theorem 2) is conditional on the calibration inequality (47); this is a substantive assumption whose failure would invalidate the benchmark comparison, but that is a correctness/robustness concern, not circularity, because the assumption does not presuppose W2S outperformance. Corollary 1 merely combines the upper and lower bounds into a sufficient condition; the condition is an inequality of explicit terms, not a restatement of the conclusion. The numerical experiments vary labeled/unlabeled sizes and overlap with prescribed feature constructions and then measure the resulting OPR; they do not fit the theoretical parameters to the observed OPR. The only self-citations (e.g., Javanmard et al. 2024) appear as background examples and are not load-bearing. The paper therefore does not reduce its central claim to its inputs by construction.

Axiom & Free-Parameter Ledger

3 free parameters · 7 axioms · 1 invented entities

The theoretical derivation is conditional on a standard set of modeling assumptions (realizability, strong convexity/smoothness, subgaussianity, identifiability of the weak model) plus an ad hoc calibration inequality in Theorem 2. No free parameters are fitted in the theory; the listed experimental constants are hand-selected tuning choices and do not affect the central claim. The only newly introduced quantity is the correlation dimension, which is a defined statistic, not a fitted knob.

free parameters (3)
  • Surrogate sharpness β = 35 (synthetic experiment)
    Hand-chosen parameter controlling the sharpness of the smoothed newsvendor surrogate in (60); not part of the theoretical claims.
  • Smoothing parameter τ = 0.5
    Used in the softplus approximation of the newsvendor cost; hand-chosen, affects experiment only.
  • W2S regularization λ_w2s = 5e-4
    Hand-chosen regularization in the experimental W2S objective (61); not used in the theoretical derivation and not justified by theory.
axioms (7)
  • domain assumption Realizability: ground-truth conditional P*(·|x) is softmax of a linear logit map η*(x) with η*1=0 and E||η*(x)||^4 < ∞.
    Assumption 1 in Section 3.1; defines the parametric family and is needed for all theorem proofs.
  • domain assumption Cost function c(a,z) is μ-strongly convex and L-smooth in a, with gradients bounded by G on W=conv(plug-in actions).
    Assumption 2 in Section 3.1; used in Lemmas 1, 2 and all subsequent bounds. The authors note it can be relaxed to convexity with technical effort.
  • domain assumption Decision set is A=R^d and outcome space is finite with K categories.
    Problem setup in Section 2.1; finite support is used throughout, whole-space is stated as relaxable.
  • ad hoc to paper Least-squares estimator bΘls is bounded in Frobenius norm by B almost surely.
    Assumption 3 in Section 3.1; explicitly a technical convenience to place bΘls in H_B, not central to the W2S effect.
  • domain assumption Features φ=[φ_s;φ_w] are subgaussian with MGF bound E exp(u^T φ) ≤ exp(½ c² u^T Σ_φ u).
    Assumption 4 in Section 3.1; needed for covariance concentration in Lemma 5 and moment bounds in Lemma 4.
  • domain assumption Weak model is identifiable: ρ_w=0, Θ*_w unique and interior; weak features bounded by B_φ.
    Assumptions 5–6 in Section 4.2; required for the MLE weak-teacher bound in Theorem 3. This is strong: the weak teacher must be correctly specified.
  • ad hoc to paper Calibration inequality: ||~w(P^s_{bΘ_s}(x)) − ~w(P^*(x))|| ≥ γ ||P^s_{bΘ_s}(x) − P^*(x)||; bounded logits; Hessian of empirical risk bounded by L_H.
    Assumptions in Theorem 2 (Eq. 47–49); imposed to make the strong-only lower bound go through. Not derived from previous assumptions and not verified experimentally.
invented entities (1)
  • correlation dimension d_{s∧w} independent evidence
    purpose: Quantifies the overlap/alignment between weak and strong feature representations; controls how much teacher error propagates to W2S decision risk.
    Defined in Definition 1 from feature covariances Σ_s, Σ_w; can be estimated from unlabeled data. Not a physical entity; the paper also proves a no-free-lunch lower bound (Lemma 12) under exact realizability, so it is not an unconstrained free parameter.

pith-pipeline@v1.3.0-alltime-deepseek · 51101 in / 16105 out tokens · 137120 ms · 2026-08-01T15:18:44.550943+00:00 · methodology

0 comments
read the original abstract

Many operational decisions rely on predictive models that estimate uncertain outcomes conditional on observable contexts. Training such models, however, often faces a fundamental data asymmetry: labeled outcomes are scarce or costly to obtain, while contextual covariates are abundant. Motivated by this data asymmetry, we develop a decision-aware weak-to-strong (W2S) framework that leverages both labeled and unlabeled data to improve contextual stochastic optimization. Specifically, we first train a weak model using limited labeled data and then use it to generate predicted outcome distributions on unlabeled contexts. These distributions provide soft supervision for training a strong model. We establish a non-asymptotic upper bound on the excess decision risk of W2S and a complementary lower bound for a strong-only benchmark. Their comparison yields explicit sufficient conditions under which W2S improves downstream decision performance. The key quantity is the correlation dimension between the weak and strong feature representations: when it is small, abundant unlabeled data reduce the effect of teacher errors along non-overlapping directions. A synthetic newsvendor experiment and a comment moderation experiment based on real-world data provide empirical evidence consistent with the theory.

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

144 extracted references · 5 canonical work pages

  1. [1]

    arXiv preprint arXiv:1804.08838 , year=

    Measuring the intrinsic dimension of objective landscapes , author=. arXiv preprint arXiv:1804.08838 , year=

    Pith/arXiv arXiv

  2. [2]

    Intrinsic dimensionality explains the effectiveness of language model fine-tuning , author=. Proceedings of the 59th annual meeting of the association for computational linguistics and the 11th international joint conference on natural language processing (volume 1: long papers) , pages=

  3. [3]

    Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Cross-Task Generalization via Natural Language Crowdsourcing Instructions , author=. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=. 2022 , publisher=

    2022

  4. [4]

    International Conference on Learning Representations , year=

    Multitask Prompted Training Enables Zero-Shot Task Generalization , author=. International Conference on Learning Representations , year=

  5. [5]

    Journal of Machine Learning Research , volume=

    Scaling Instruction-Finetuned Language Models , author=. Journal of Machine Learning Research , volume=

  6. [6]

    SIAM Journal on Mathematics of Data Science , volume=

    Why are big data matrices approximately low rank? , author=. SIAM Journal on Mathematics of Data Science , volume=. 2019 , publisher=

    2019

  7. [7]

    International conference on machine learning , pages=

    More than a toy: Random matrix models predict how real-world neural representations generalize , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  8. [8]

    Advances in Neural Information Processing Systems , volume=

    Sketchy moment matching: Toward fast and provable data selection for finetuning , author=. Advances in Neural Information Processing Systems , volume=

  9. [9]

    Advances in neural information processing systems , volume=

    Neural tangent kernel: Convergence and generalization in neural networks , author=. Advances in neural information processing systems , volume=

  10. [10]

    Conference on Learning Theory , pages=

    Kernel and rich regimes in overparametrized models , author=. Conference on Learning Theory , pages=. 2020 , organization=

    2020

  11. [13]

    Advances in Neural Information Processing Systems , volume=

    Training compute-optimal large language models , author=. Advances in Neural Information Processing Systems , volume=. 2022 , doi=

    2022

  12. [14]

    Advances in neural information processing systems , volume=

    Language models are few-shot learners , author=. Advances in neural information processing systems , volume=

  13. [15]

    Proceedings of the 41st International Conference on Machine Learning , pages=

    Weak-to-strong generalization: eliciting strong capabilities with weak supervision , author=. Proceedings of the 41st International Conference on Machine Learning , pages=

  14. [16]

    Proceedings of the 42nd International Conference on Machine Learning , series=

    Discrepancies are virtue: Weak-to-strong generalization through lens of intrinsic dimension , author=. Proceedings of the 42nd International Conference on Machine Learning , series=. 2025 , publisher=

    2025

  15. [17]

    The Fourteenth International Conference on Learning Representations , year=

    Does Weak-to-strong Generalization Happen under Spurious Correlations? , author=. The Fourteenth International Conference on Learning Representations , year=

  16. [18]

    Advances in neural information processing systems , volume=

    Theoretical analysis of weak-to-strong generalization , author=. Advances in neural information processing systems , volume=

  17. [19]

    International Conference on Machine Learning , pages=

    A theory of label propagation for subpopulation shift , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  18. [20]

    Patterns , volume=

    Unleashing the potential of prompt engineering for large language models , author=. Patterns , volume=. 2025 , publisher=

    2025

  19. [21]

    Advances in neural information processing systems , volume=

    Quantifying the gain in weak-to-strong generalization , author=. Advances in neural information processing systems , volume=

  20. [22]

    arXiv preprint arXiv:2402.03749 , year=

    Vision superalignment: Weak-to-strong generalization for vision foundation models , author=. arXiv preprint arXiv:2402.03749 , year=

    Pith/arXiv arXiv

  21. [23]

    arXiv preprint arXiv:2402.15505 , year=

    Co-supervised learning: Improving weak-to-strong generalization with hierarchical mixture of experts , author=. arXiv preprint arXiv:2402.15505 , year=

    Pith/arXiv arXiv

  22. [25]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Debate helps weak-to-strong generalization , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  23. [26]

    ACM Transactions on Intelligent Systems and Technology , volume=

    Survey on knowledge distillation for large language models: methods, evaluation, and application , author=. ACM Transactions on Intelligent Systems and Technology , volume=. 2025 , publisher=

    2025

  24. [27]

    Computer-Aided Civil and Infrastructure Engineering , volume=

    Weakly-supervised structural component segmentation via scribble annotations , author=. Computer-Aided Civil and Infrastructure Engineering , volume=. 2025 , publisher=

    2025

  25. [28]

    Proceedings of the Eleventh Annual Conference on Computational Learning Theory , pages=

    Combining labeled and unlabeled data with co-training , author=. Proceedings of the Eleventh Annual Conference on Computational Learning Theory , pages=

  26. [29]

    Proceedings of the 20th International Conference on Machine Learning , pages=

    Semi-supervised learning using Gaussian fields and harmonic functions , author=. Proceedings of the 20th International Conference on Machine Learning , pages=

  27. [30]

    Advances in Neural Information Processing Systems , volume=

    Semi-supervised learning by entropy minimization , author=. Advances in Neural Information Processing Systems , volume=

  28. [31]

    ICML Workshop on Challenges in Representation Learning , year=

    Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks , author=. ICML Workshop on Challenges in Representation Learning , year=

  29. [32]

    Advances in Neural Information Processing Systems , volume=

    FixMatch: Simplifying semi-supervised learning with consistency and confidence , author=. Advances in Neural Information Processing Systems , volume=

  30. [33]

    , author=

    Does Unlabeled Data Provably Help? Worst-case Analysis of the Sample Complexity of Semi-Supervised Learning. , author=. COLT , pages=

  31. [34]

    Advances in neural information processing systems , volume=

    Unlabeled data improves adversarial robustness , author=. Advances in neural information processing systems , volume=

  32. [35]

    International Conference on Learning Representations , year=

    Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data , author=. International Conference on Learning Representations , year=

  33. [36]

    IEEE transactions on knowledge and data engineering , volume=

    A survey on deep semi-supervised learning , author=. IEEE transactions on knowledge and data engineering , volume=. 2022 , publisher=

    2022

  34. [38]

    Operations Research , volume=

    Data-driven optimization: A reproducing kernel hilbert space approach , author=. Operations Research , volume=. 2022 , publisher=

    2022

  35. [39]

    International conference on neural information processing , pages=

    Assessing the performance of deep learning algorithms for newsvendor problem , author=. International conference on neural information processing , pages=. 2017 , organization=

    2017

  36. [40]

    European Journal of Operational Research , volume=

    A data-driven newsvendor problem: From data to decision , author=. European Journal of Operational Research , volume=. 2019 , publisher=

    2019

  37. [41]

    predict, then optimize

    Smart “predict, then optimize” , author=. Management Science , volume=. 2022 , publisher=

    2022

  38. [42]

    Advances in neural information processing systems , volume=

    Generalization bounds in the predict-then-optimize framework , author=. Advances in neural information processing systems , volume=

  39. [43]

    International conference on machine learning , pages=

    Decision trees for decision-making under the predict-then-optimize framework , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  40. [45]

    Tutorials in operations research: emerging and impactful topics in operations , pages=

    Integrating prediction/estimation and optimization with applications in operations management , author=. Tutorials in operations research: emerging and impactful topics in operations , pages=. 2022 , publisher=

    2022

  41. [46]

    Operations Research , volume=

    Integrated conditional estimation-optimization , author=. Operations Research , volume=. 2025 , doi=

    2025

  42. [47]

    Management Science , volume=

    From predictive to prescriptive analytics , author=. Management Science , volume=. 2020 , publisher=

    2020

  43. [48]

    Available at SSRN 3623006 , year=

    Decision-Driven Regularization: A Blended Model for Learning and Optimization , author=. Available at SSRN 3623006 , year=

  44. [49]

    Advances in neural information processing systems , volume=

    Task-based end-to-end model learning in stochastic optimization , author=. Advances in neural information processing systems , volume=

  45. [50]

    Operations Research , volume=

    The big data newsvendor: Practical insights from machine learning , author=. Operations Research , volume=. 2019 , publisher=

    2019

  46. [51]

    Computational Management Science , volume=

    Predictive stochastic programming , author=. Computational Management Science , volume=. 2022 , publisher=

    2022

  47. [52]

    Manufacturing & Service Operations Management , volume=

    On Data-Driven Prescriptive Analytics with Side Information: A Regularized Nadaraya--Watson Approach , author=. Manufacturing & Service Operations Management , volume=. 2026 , doi=

    2026

  48. [53]

    Available at SSRN 5606310 , year=

    Feature-Rich, Data-Private: A Sparse Learning Framework for the High-Dimensional Newsvendor , author=. Available at SSRN 5606310 , year=

  49. [54]

    Manufacturing & Service Operations Management , volume=

    Analytics for an Online Retailer: Demand Forecasting and Price Optimization , author=. Manufacturing & Service Operations Management , volume=. 2016 , doi=

    2016

  50. [55]

    Mathematics of Operations Research , volume=

    Nonparametric Learning Algorithms for Joint Pricing and Inventory Control with Lost Sales and Censored Demand , author=. Mathematics of Operations Research , volume=. 2021 , doi=

    2021

  51. [56]

    Management Science , volume=

    Utility Fairness in Contextual Dynamic Pricing with Demand Learning , author=. Management Science , volume=. 2026 , doi=

    2026

  52. [57]

    Computational Management Science , volume=

    Likelihood Robust Optimization for Data-Driven Problems , author=. Computational Management Science , volume=. 2016 , doi=

    2016

  53. [58]

    Management Science , volume=

    Dynamic Pricing and Inventory Control with Fixed Ordering Cost and Incomplete Demand Information , author=. Management Science , volume=. 2022 , doi=

    2022

  54. [59]

    Innovative Technology at the Interface of Finance and Operations , pages=

    Applied Machine Learning in Operations Management , author=. Innovative Technology at the Interface of Finance and Operations , pages=. 2022 , doi=

    2022

  55. [60]

    Operations Research , volume=

    Offline Multi-Action Policy Learning: Generalization and Optimization , author=. Operations Research , volume=. 2023 , doi=

    2023

  56. [61]

    Operations Research , volume=

    Technical Note--Nonparametric Data-Driven Algorithms for Multiproduct Inventory Systems with Censored Demand , author=. Operations Research , volume=. 2016 , doi=

    2016

  57. [62]

    Management Science , volume=

    Policy Learning with Adaptively Collected Data , author=. Management Science , volume=. 2023 , doi=

    2023

  58. [63]

    Management Science , year=

    Collaborative Learning and Decision Making on Pricing and Recommendation: A Simple Framework for Planning , author=. Management Science , year=

  59. [64]

    Management Science , volume=

    Contextual Learning with Online Convex Optimization: Theory and Application to Medical Decision-Making , author=. Management Science , volume=. 2025 , doi=

    2025

  60. [65]

    INFORMS Journal on Computing , year=

    Contextual Stochastic Vehicle Routing with Time Windows , author=. INFORMS Journal on Computing , year=

  61. [66]

    European Journal of Operational Research , volume=

    A survey of contextual optimization methods for decision-making under uncertainty , author=. European Journal of Operational Research , volume=. 2025 , publisher=

    2025

  62. [67]

    International Conference on Algorithmic Learning Theory , pages=

    A vector-contraction inequality for rademacher complexities , author=. International Conference on Algorithmic Learning Theory , pages=. 2016 , organization=

    2016

  63. [69]

    2018 , publisher=

    High-dimensional probability: An introduction with applications in data science , author=. 2018 , publisher=

    2018

  64. [71]

    2019 , publisher=

    High-Dimensional Statistics: A Non-Asymptotic Viewpoint , author=. 2019 , publisher=

    2019

  65. [72]

    Advances in Neural Information Processing Systems , volume=

    Fair and welfare-efficient constrained multi-matchings under uncertainty , author=. Advances in Neural Information Processing Systems , volume=

  66. [73]

    Management Science , volume=

    Feature-based dynamic pricing , author=. Management Science , volume=. 2020 , publisher=

    2020

  67. [75]

    The Review of Financial Studies , volume=

    Empirical asset pricing via machine learning , author=. The Review of Financial Studies , volume=. 2020 , publisher=

    2020

  68. [76]

    Transportation Research Part C: Emerging Technologies , volume=

    Data-driven robust optimization for contextual vehicle rebalancing in on-demand ride services under demand uncertainty , author=. Transportation Research Part C: Emerging Technologies , volume=. 2023 , publisher=

    2023

  69. [79]

    Seventeenth Symposium on Usable Privacy and Security (SOUPS 2021) , pages=

    Designing Toxic Content Classification for a Diversity of Perspectives , author=. Seventeenth Symposium on Usable Privacy and Security (SOUPS 2021) , pages=. 2021 , month=

    2021

  70. [81]

    Companion Proceedings of The 2019 World Wide Web Conference , pages=

    Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification , author=. Companion Proceedings of The 2019 World Wide Web Conference , pages=

    2019

  71. [82]

    2019 , publisher=

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , booktitle=. 2019 , publisher=

    2019

  72. [83]

    2019 , eprint=

    Sanh, Victor and Debut, Lysandre and Chaumond, Julien and Wolf, Thomas , booktitle=. 2019 , eprint=

    2019

  73. [84]

    The sign estimator: Llm alignment in the face of choice heterogeneity

    Ali Aouad, Aymane El Gadarri, and Vivek F Farias. The sign estimator: Llm alignment in the face of choice heterogeneity. arXiv preprint arXiv:2510.23965v2, 2025

    arXiv 2025

  74. [85]

    The big data newsvendor: Practical insights from machine learning

    Gah-Yi Ban and Cynthia Rudin. The big data newsvendor: Practical insights from machine learning. Operations Research, 67 0 (1): 0 90--108, 2019

    2019

  75. [86]

    Zhang, and Heng Zhang

    Hamsa Bastani, Dennis J. Zhang, and Heng Zhang. Applied machine learning in operations management. In Innovative Technology at the Interface of Finance and Operations, pages 189--222. Springer, 2022. doi:10.1007/978-3-030-75729-8_7

    work page doi:10.1007/978-3-030-75729-8_7 2022

  76. [87]

    Does unlabeled data provably help? worst-case analysis of the sample complexity of semi-supervised learning

    Shai Ben-David, Tyler Lu, and D \'a vid P \'a l. Does unlabeled data provably help? worst-case analysis of the sample complexity of semi-supervised learning. In COLT, pages 33--44, 2008

    2008

  77. [88]

    Data-driven optimization: A reproducing kernel hilbert space approach

    Dimitris Bertsimas and Nihal Koduri. Data-driven optimization: A reproducing kernel hilbert space approach. Operations Research, 70 0 (1): 0 454--471, 2022

    2022

  78. [89]

    Combining labeled and unlabeled data with co-training

    Avrim Blum and Tom Mitchell. Combining labeled and unlabeled data with co-training. In Proceedings of the Eleventh Annual Conference on Computational Learning Theory, pages 92--100, 1998

    1998

  79. [90]

    Nuanced metrics for measuring unintended bias with real data for text classification

    Daniel Borkan, Lucas Dixon, Jeffrey Sorensen, Nithum Thain, and Lucy Vasserman. Nuanced metrics for measuring unintended bias with real data for text classification. In Companion Proceedings of The 2019 World Wide Web Conference, pages 491--500, 2019

    2019

  80. [91]

    Weak-to-strong generalization: eliciting strong capabilities with weak supervision

    Collin Burns, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold Aschenbrenner, Yining Chen, Adrien Ecoffet, Manas Joglekar, Jan Leike, et al. Weak-to-strong generalization: eliciting strong capabilities with weak supervision. In Proceedings of the 41st International Conference on Machine Learning, pages 4971--5012, 2024

    2024

Showing first 80 references.