Pith.

Machine Learning · arXiv:2611.17059 v1 · submitted 12 Nov 2026 · 125 references

Capacity-Constrained Delayed Feedback for Online Conformal Risk Control

Bowen Wang, Osvaldo Simeone

The paper says

“Online conformal risk control (OCRC) enables reliable uncertainty quantification in machine learning by converting model outputs into prediction sets with long-term risk guarantees.”

more of the abstract

These guarantees, however, hinge on the assumption that the loss of each round is revealed immediately and losslessly. In practice, feedback may be delayed, and limited resources may prevent tracking all pending predictions, causing some feedback to be lost. To address this challenge, we develop capacity-aware OCRC (CA-OCRC), a framework that jointly determines how to allocate limited tracking feedback capacity and how to construct prediction sets. CA-OCRC is based on a randomized tracking policy that samples a proxy delay to decide whether, and for how long, feedback for the current round is tracked. CA-OCRC updates the prediction sets via inverse probability weighting, compensating for the possibility of feedback lost due to delays and limited tracking capacity. We analyze risk bounds for both arbitrary and stochastic delays, revealing a trade-off between receiving more feedback and avoiding capacity saturation. Building on these bounds, we optimize the proxy delay distribution, showing scaling laws for the tracking capacity that ensure the same long-term risk as OCRC with unlimited feedback. Experiments on medical imaging, natural language processing, computer vision, and wireless communications show the effectiveness of CA-OCRC as compared to existing OCRC mechanisms.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up, with conditions. A solid capacity-aware OCRC framework whose arbitrary-delay results are rigorous, but the stochastic-delay scaling law in Corollary 2 has a real proof gap in the K > E[dt] regime. This line is wrong

1 major objection · 4 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 125 references say.

The paper says, at Appendix E-B, Corollary 2 (Eq. (38))

The review did not quote this passage.

Pith says: Objection 1 of 1 · major

The proof of Corollary 2 bounds Theorem 2 with the constant proxy law F(0)=1 and F(a)=p for a>=1, and chooses p = K/(e E[dt]) (E[dt]/T)^{1/(K+1)}. It asserts that p<=1 and p E[dt]<=K 'since K<=E[dt]<=T'. However, Corollary 2's hypotheses do not include K<=E[dt] or E[dt]<=T.

more of objection 1

When K>E[dt], the chosen p can exceed 1; for example, with K=20, E[dt]=1 and T=10^6, p is approximately 3.8, which violates the feasible-survival-function constraints in (35c) and makes the inverse weight 1/F(dt) ill-defined. Because the reduction 'it suffices to bound (32) for ... F(a)=p' requires p to be a feasible proxy distribution, the proof does not establish the advertised mean-delay scaling for all K. The feasible choice p=1 gives, through Theorem 2, a bound of order (K+E[dt])/T together with a non-vanishing saturation term, not the claimed C_alpha (E[dt]/T)^{K/(K+1)} in this regime. Corollary 2 therefore needs an explicit hypothesis such as K<=E[dt] (and E[dt]<=T, also used), or a separate large-capacity argument; the 'any capacity K' statements in Sec. I-C and Sec. VIII must be revised accordingly. For fixed K and E[dt] the chosen p eventually becomes feasible for sufficiently large T, so the asymptotic rate may be recoverable, but that qualifier is absent from the theorem as stated.

4 minor objections

The paper says, at Sec. II-C, Eq. (10)

Pith says: minor objection 1 of 4

The proof of Corollary 2 also divides by E[dt] when defining p; the degenerate case E[dt]=0, where all delays are zero, should be excluded or handled separately in the statement.

The paper says, at Appendix E, Proposition 2

Pith says: minor objection 2 of 4

The closed-form solution (E-11) for the Bregman isotonic regression step is presented very tersely; a brief derivation or an explicit reference for the isotonic solution formula would improve reproducibility.

show 2 more

The paper says, at Sec. I-A

Pith says: minor objection 3 of 4

There is a typo in the motivation paragraph: 'it may lost' should read 'it may be lost.'

The paper says, at Sec. VI-A, Eq. (33)

Pith says: minor objection 4 of 4

The definition of R_F,G uses the conditional expectation E_G[dt | dt <= dmax]; the text should state explicitly that the theorem is used only when q_G(dmax)>0, although in the main setting this follows from dmax being the maximum possible delay.

What would settle it

Numerical check of the proof, not of the theorem: take K=20, E[dt]=1, dmax=T=10^6, α=0.1. First verify that the p in Appendix E-B is ≈3.8, violating 1=F(-1)≥F(0)≥F(1). Then search the feasible family (36) from Proposition 2 (varying θ so that ℓFθ,G≤K) and evaluate the Theorem 2 bound (32) at its optimum. If the best feasible bound exceeds Bη/T + 2ᾱ/T + Cα(E[dt]/T)^{K/(K+1)}, Corollary 2 as stated needs the extra hypothesis K≤E[dt] or a new argument for K>E[dt]; if some feasible F matches the bound, restore the proof with that F.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence moderate
  2. Skeptic partly agrees 3 min 37 s · 37,370 tokens read, 31,143 written
  3. Editor wrote the verdict above 41 s · 38,981 tokens read, 595 written
  4. Referee major revision: 1 major objection, 4 minor 1 min 58 s · 37,859 tokens read, 18,642 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

The paper delivers something genuinely useful: the first risk-control guarantees for online conformal prediction under both delayed feedback and a finite tracking budget. The arbitrary-delay analysis (Theorem 1, Corollary 1) is coherent and, as far as I can tell, correct. The threshold telescoping, the saturation bound via Chernoff, and the scaling law K = O(log(T/dmax)) all hang together. The two-level inverse probability weighting is a natural and clean extension of prior intermittent-feedback work, and the paper is honest about where distribution-aware designs need G to be known and i.i.d.

The soft spot is real and it is in Corollary 2. The proof of that corollary chooses an admission probability p = K/(e E[dt]) (E[dt]/T)1/(K+1) and claims p <= 1 using an assumption K <= E[dt] that never appears in the statement. When K > E[dt] — which nothing rules out, and which is actually the practically interesting case when delays are short relative to capacity — that p exceeds 1, so it is not a valid survival function and the inverse weights 1/F(dt) are not probabilities. The paper's own underload constraint (35b) is not what is being violated; the feasibility of the specific p is. Without that step, the stochastic scaling law O((E[dt]/T)K/(K+1)) is not established for K > E[dt]. It is a gap in one corollary's proof, not in Theorem 2 or in the arbitrary-delay results, and the paper's own worst-case Theorem 1 still covers that regime via dmax. But as written, the claim in the abstract and Sec. VI-C that the mean-delay rate holds for any capacity K is unsupported.

The experiments are consistent with the qualitative story — more capacity converges faster, worst-case design is conservative, the F=G heuristic saturates — but there is no code or data, so independent verification is limited. That is a minor issue, not a blocker. The paper's limitation statements about correlated delays and unknown G are accurate and appropriately scoped.

This deserves a serious referee. The fix is straightforward: either add K <= E[dt] to Corollary 2's hypotheses, or give a separate argument for the large-capacity regime. I would accept to review it and my verdict would be conditional on that correction.

The strongest objection

Corollary 2 is the paper's stochastic-delay scaling law: risk gap ≤ Bη/T + 2ᾱ/T + Cα(E[dt]/T)^{K/(K+1)}. Its proof in Appendix E-B bounds Theorem 2 with the constant proxy law F(0)=1, F(a)=p for a∈{1,...,dmax}, choosing p = K/(e E[dt]) (E[dt]/T)^{1/(K+1)}, and asserts p≤1 "since K≤E[dt]≤T". But Corollary 2's hypotheses do not include K≤E[dt]. When K>E[dt], p can exceed 1; e.g., K=20, E[dt]=1, T=10^6 gives p≈3.8. A survival function must satisfy 1=F(-1)≥F(0)≥...≥F(dmax)≥0, as in (27c)/(35c), so this p does not define a valid distribution and the inverse weights 1/F(dt) are not well-defined probabilities. The feasible fallback F≡1 gives, through Theorem 2, only the K QF,G term in (33), i.e., a K/T-type bound, which is not shown to lie below Cα(E[dt]/T)^{K/(K+1)} in this regime. Hence the advertised "any capacity K" mean-delay rate is not proved for K>E[dt]; a condition such as K≤E[dt], or a separate large-capacity argument, is needed. This is a proof gap, not an attack on Theorem 1/Corollary 1.

I read the paper as establishing a worst-case guarantee (Theorem 1, Corollary 1) and a distribution-dependent refinement (Theorem 2, Corollary 2). The arbitrary-delay chain is solid: the saturation Chernoff bound, the K+1 threshold-range argument, and the Bernoulli optimization in Proposition 1 are internally consistent, and the experiments corroborate the qualitative trade-off. The load-bearing soft spot is Corollary 2's proof, not the i.i.d./known-G assumption, which the paper acknowledges in Sec. VII-D. Appendix E-B constructs p>1 in the K>E[dt] regime, violating the survival-function monotonicity constraints. Thus the mean-delay scaling law for "any capacity K" is not proven. This matches the reader's conditional verdict and the feasibility concern in their rationale, so I recommend no change: UNCHANGED. A concrete numerical regime can settle whether the statement needs a condition or whether a feasible F in Proposition 2's family attains the bound.

The referee's report

If the results hold as stated, the paper makes a useful contribution: it extends the delayed-feedback literature for online conformal prediction from the unlimited-capacity regime to a realistic limited-tracking-capacity regime, and it gives an explicit rate trade-off between capacity and risk convergence. The arbitrary-delay part appears carefully derived: Theorem 1 combines a threshold-telescoping argument with a Chernoff-type saturation bound, and Corollary 1 follows from a feasible Bernoulli proxy law. The stochastic-delay refinement, replacing dmax by E[dt], is natural and, if corrected, would strengthen the paper considerably. The experiments cover diverse tasks and compare against reasonable baselines, and the authors are explicit that the correlated-delay experiments are covered only by the worst-case arbitrary-delay guarantee. However, one load-bearing proof step in the stochastic scaling law is not supported as written, so the paper needs revision before the advertised claims can be accepted.

Pith's reading of the paper (a machine's summary, not the authors')

Online conformal prediction turns a model's scores into sets that contain the truth with a chosen risk level. In the online setting, the threshold is updated using the loss of each round. This paper considers the practical case where the true loss arrives only after a delay and where the system can track only K pending rounds at once. If the tracking buffer is full, some feedback is lost permanently.

Core claim. The central claim, Theorem 1 and Corollary 1, is that with capacity K and a proxy-delay distribution satisfying the underload condition, the expected risk gap of CA-OCRC is at most B*eta/T + (2e+1)*(dmax/T)^(K/(K+1)), so for fixed K the gap vanishes when dmax = o(T), and a capacity K of order log(T/dmax) recovers the O(dmax/T) convergence rate of unlimited-capacity OCRC. The stochastic-delay version, Corollary 2, replaces dmax with the mean delay E[dt].

The premise it rests on. The stochastic-delay results (Theorem 2, Proposition 2, Corollary 2) assume the delay distribution G is known exactly and the delays are i.i.d.; CA-OCRC's inverse probability weights and optimized proxy distribution use G directly in Eqs. (22), (23), and (36). If G is misspecified or delays are correlated, the distribution-aware guarantee is not valid and only the worst-case dmax bound in Theorem 1 applies, as the authors note in Sec. VII-D.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Online conformal risk control (OCRC) enables reliable uncertainty quantification in machine learning by converting model outputs into prediction sets with long-term risk guarantees. These guarantees, however, hinge on the assumption that the loss of each round is revealed immediately and losslessly. In practice, feedback may be delayed, and limited resources may prevent tracking all pending predictions, causing some feedback to be lost. To address this challenge, we develop capacity-aware OCRC (CA-OCRC), a framework that jointly determines how to allocate limited tracking feedback capacity and how to construct prediction sets. CA-OCRC is based on a randomized tracking policy that samples a proxy delay to decide whether, and for how long, feedback for the current round is tracked. CA-OCRC updates the prediction sets via inverse probability weighting, compensating for the possibility of feedback lost due to delays and limited tracking capacity. We analyze risk bounds for both arbitrary and stochastic delays, revealing a trade-off between receiving more feedback and avoiding capacity saturation. Building on these bounds, we optimize the proxy delay distribution, showing scaling laws for the tracking capacity that ensure the same long-term risk as OCRC with unlimited feedback. Experiments on medical imaging, natural language processing, computer vision, and wireless communications show the effectiveness of CA-OCRC as compared to existing OCRC mechanisms.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-14

The authors' main claim, as the classifier picked it out 0.55

“Experiments on medical imaging, natural language processing, computer vision, and wireless communications show the effectiveness of CA-OCRC as compared to existing OCRC mechanisms.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.43 · human participants 0.16 · controlled comparison 0.73 · randomized assignment 0.20 · states quantitative result 0.22 · compares with prior results 0.91 · mentions code or data availability 0.02

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Online conformal risk control (OCRC) enables reliable uncertainty quantification in machine learning by converting model outputs into prediction sets with long-term risk guarantees. background 1.00 0.08 0.05 asserted 1.00
These guarantees, however, hinge on the assumption that the loss of each round is revealed immediately and losslessly. background 0.59 0.09 0.02 asserted 1.00
In practice, feedback may be delayed, and limited resources may prevent tracking all pending predictions, causing some feedback to be lost. background 0.60 0.07 0.02 hedged possibility 0.95
To address this challenge, we develop capacity-aware OCRC (CA-OCRC), a framework that jointly determines how to allocate limited tracking feedback capacity and how to construct prediction sets. contribution 0.97 0.95 0.15 asserted 1.00
CA-OCRC is based on a randomized tracking policy that samples a proxy delay to decide whether, and for how long, feedback for the current round is tracked. method 0.98 0.90 0.05 asserted 1.00
CA-OCRC updates the prediction sets via inverse probability weighting, compensating for the possibility of feedback lost due to delays and limited tracking capacity. method 0.99 0.86 0.09 asserted 0.99
We analyze risk bounds for both arbitrary and stochastic delays, revealing a trade-off between receiving more feedback and avoiding capacity saturation. result 0.75 0.95 0.15 asserted 1.00
Building on these bounds, we optimize the proxy delay distribution, showing scaling laws for the tracking capacity that ensure the same long-term risk as OCRC with unlimited feedback. result 0.53 0.96 0.39 asserted 1.00
Experiments on medical imaging, natural language processing, computer vision, and wireless communications show the effectiveness of CA-OCRC as compared to existing OCRC mechanisms. result 0.95 0.95 0.37 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.55

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-14: 1 request, 6,890 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (13)
Figure 1 of the paper
Figure 1. Illustration of sequential prediction with delayed feedback under a finite tracking capacity. (a) Operations within round
Figure 2 of the paper
Figure 2. FNR as a function of the time horizon T for the tumor segmentation task with ν = 4, θ = 8, ρ = 0, dmax = 150, and target level α = 0.1 (dashed line), for capacities (a) K = 1, (b) K = 2, and (c) K = 4. The results are averaged over 103 independent trials, with the error bars denoting ±1 standard deviation. The mean and variance of the delays are νθ and νθ2 , respectively. For ρ = 0, we set Ut = 0, and thus the delays are independent and identically distributed (i.i.d.) as in Assumption 1(b), while for ρ > 0, they are identically distributed but temporally correlated, and realization…
Figure 3 of the paper
Figure 3. (a) FNR, (b) set size, and (c) tracking load as a function of the capacity
Figure 4 of the paper
Figure 4. (a) FNR, (b) set size, and (c) tracking load as a function of the shape
Figure 5 of the paper
Figure 5. F1 risk as a function of the time horizon
Figure 6 of the paper
Figure 6. (a) F1 risk, (b) set size, and (c) tracking load as a function of the capacity
Figure 7 of the paper
Figure 7. (a) F1 risk, (b) set size, and (c) tracking load as a function of the scale parameter
Figure 8 of the paper
Figure 8. (a) Risk gap, (b) set size, and (c) tracking load of CA-OCRC with a fixed admission probability
Figure 9 of the paper
Figure 9. Hierarchical risk as a function of the time horizon
Figure 10 of the paper
Figure 10. (a) Hierarchical risk, (b) set size, and (c) tracking load as a function of the capacity
Figure 11 of the paper
Figure 11. (a) Hierarchical risk, (b) set size, and (c) tracking load as a function of the correlation time
Figure 12 of the paper
Figure 12. SNR loss as a function of the time horizon
Figure 13 of the paper
Figure 13. (a) SNR loss, (b) set size, and (c) tracking load as a function of the capacity

Figures are the authors', reproduced from the paper.

References (125)
  1. [1] H. Qiu, E. Esposito, and M. Zhang, ``Exploiting curvature in online convex optimization with delayed feedback,'' in Proc. Int. Conf. Mach. Learn., 2025, pp. 50\,448--50\,479 on Pith
  2. [2] X. Cao, J. Zhang, and H. V. Poor, ``Constrained online convex optimization with feedback delays,'' IEEE Trans. Autom. Control , vol. 66, no. 11, pp. 5049--5064, 2020 on Pith
  3. [3] B. Wang, M. Zecchin, and O. Simeone, ``Distributed dynamic associative memory via online convex optimization,'' IEEE J. Sel. Areas Commun. , vol. 44, pp. 5872--5887, 2026 on Pith
  4. [4] ------, ``Mirror online conformal prediction with intermittent feedback,'' IEEE Signal Process. Lett. , vol. 32, pp. 2888--2892, 2025 on Pith
  5. [5] M. Zhao, H. Admoni, R. Simmons, A. Ramdas, and A. Bajcsy, ``Conformalized interactive imitation learning: Handling expert shift and intermittent feedback,'' in Proc. Int. Conf. Learn. Represent., 2025, pp. 9238--9251 on Pith
  6. [6] R. Vershynin, High-Dimensional Probability: An Introduction with Applications in Data Science. 1em plus 0.5em minus 0.4em Cambridge University Press, 2018 on Pith
  7. [7] R. Durrett, Probability: theory and examples. 1em plus 0.5em minus 0.4em Cambridge university press, 2019, vol. 49 on Pith
  8. [8] W. Wisniewski, D. Lindsay, and S. Lindsay, ``Application of conformal prediction interval estimations to market makers’ net positions,'' in Proc. Symp. Conformal Probabilistic Prediction Appl., 2020, pp. 285--301 on Pith
  9. [9] J. Lekeufack, A. N. Angelopoulos et al., ``Conformal decision theory: Safe autonomous decisions from imperfect predictions,'' in Proc. IEEE Int. Conf. Robot. Automat., 2024, pp. 11\,668--11\,675 on Pith
  10. [10] O. Simeone, S. Park, and M. Zecchin, ``Conformal calibration: Ensuring the reliability of black-box AI in wireless systems,'' IEEE Commun. Mag. , vol. 64, no. 8, pp. 179--187, 2026 on Pith
  11. [11] Y.-J. Park, C. Sobolewski, and N. Azizan, ``Uncertainty quantification in detection transformers: Object-level calibration and image-level reliability,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 48, no. 11, pp. 13\,775--13\,788, 2026 on Pith
  12. [12] K. Wang, F. Cuzzolin, K. Shariatmadar, D. Moens, and H. Hallez, ``A review of uncertainty representation and quantification in neural networks,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 48, no. 3, pp. 2476--2495, 2026 on Pith
  13. [13] O. Simeone and Y. Romano, ``Uncertainty quantification and data efficiency in AI : An information-theoretic perspective,'' IEEE J. Sel. Areas Inf. Theory, vol. 7, pp. 290--301, 2026 on Pith
  14. [15] S. Park, K. M. Cohen, and O. Simeone, ``Few-shot calibration of set predictors via meta-learned cross-validation-based conformal prediction,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 46, no. 1, pp. 280--291, 2024 on Pith
  15. [16] A. N. Angelopoulos, S. Bates, A. Fisch, L. Lei, and T. Schuster, ``Conformal risk control,'' in Proc. Int. Conf. Learn. Represent., 2024, pp. 55\,198--55\,218 on Pith
  16. [17] B.-S. Einbinder, L. Ringel, and Y. Romano, ``Semi-supervised risk control via prediction-powered inference,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 47, no. 12, pp. 11\,080--11\,090, 2025 on Pith
  17. [18] V. Blot, A. N. Angelopoulos, M. Jordan, and N. J.-B. Brunel, ``Automatically adaptive conformal risk control,'' in Proc. Int. Conf. Artif. Intell. Statist., 2025 on Pith
  18. [19] I. Gibbs and E. Candes, ``Adaptive conformal inference under distribution shift,'' in Proc. Adv. Neural Inf. Process. Syst., vol. 34, 2021, pp. 1660--1672 on Pith
  19. [20] A. N. Angelopoulos, R. Barber, and S. Bates, ``Online conformal prediction with decaying step sizes,'' in Proc. Int. Conf. Mach. Learn., 2024, pp. 1616--1630 on Pith
  20. [21] C. Xu and Y. Xie, ``Conformal prediction for time series,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 45, no. 10, pp. 11\,575--11\,587, 2023 on Pith
show 80 more
  1. [22] S. Feldman, L. Ringel, S. Bates, and Y. Romano, ``Achieving risk control in online learning settings,'' Trans. Mach. Learn. Res., 2023 on Pith
  2. [23] M. Zecchin and O. Simeone, ``Localized adaptive risk control,'' in Proc. Adv. Neural Inf. Process. Syst., vol. 37, 2024, pp. 8165--8192 on Pith
  3. [24] M. Zhu, M. Zecchin, S. Park, C. Guo, C. Feng, P. Popovski, and O. Simeone, ``Conformal distributed remote inference in sensor networks under reliability and communication constraints,'' IEEE Trans. Signal Process. , vol. 73, pp. 1485--1500, 2025 on Pith
  4. [25] F. Binucci, O. Simeone, and P. Banelli, ``Conformal Lyapunov optimization: Optimal resource allocation under deterministic reliability constraints,'' IEEE Trans. Signal Process. , vol. 73, pp. 4521--4538, 2025 on Pith
  5. [26] C. L. Cifra, D. F. Sittig, and H. Singh, ``Bridging the feedback gap: a sociotechnical approach to informing clinicians of patients subsequent clinical course and outcomes,'' BMJ Qual. Saf., vol. 30, no. 7, pp. 591--597, 2021 on Pith
  6. [27] T. Kwiatkowski et al., `` Natural Questions : A benchmark for question answering research,'' Trans. Assoc. Comput. Linguist., vol. 7, pp. 452--466, 2019 on Pith
  7. [28] B. Csaba, W. Zhang, M. M \"u ller, S.-N. Lim, M. Elhoseiny, P. H. Torr, and A. Bibi, ``Label delay in online continual learning,'' in Proc. Adv. Neural Inf. Process. Syst., vol. 37, 2024, pp. 119\,976--120\,012 on Pith
  8. [29] Q. Pham, C. Liu, and S. C. Hoi, ``Continual learning, fast and slow,'' IEEE Trans. Pattern Anal. Mach. Intell. , vol. 46, no. 1, pp. 134--149, 2023 on Pith
  9. [31] O. Chapelle, ``Modeling delayed feedback in display advertising,'' in Proc. ACM SIGKDD Int. Conf. Knowl. Discov. Data Min., 2014, pp. 1097--1105 on Pith
  10. [32] J. Hallberg Szabadv\' a ry, ``Adaptive conformal inference for multi-step ahead time-series forecasting online,'' in Proc. Symp. Conformal Probabilistic Prediction Appl., 2024, pp. 250--263 on Pith
  11. [35] B. Wang, M. Zecchin, and O. Simeone, ``Mirror online conformal prediction with intermittent feedback,'' IEEE Signal Process. Lett. , vol. 32, pp. 2888--2892, 2025 on Pith
  12. [37] S. Shalev-Shwartz, ``Online learning and online convex optimization,'' Found. Trends Mach. Learn., vol. 4, no. 2, pp. 107--194, 2025 on Pith
  13. [38] M. Zinkevich, ``Online convex programming and generalized infinitesimal gradient ascent,'' in Proc. Int. Conf. Mach. Learn., 2003, pp. 928--936 on Pith
  14. [40] K. Quanrud and D. Khashabi, ``Online learning with adversarial delays,'' in Proc. Adv. Neural Inf. Process. Syst., vol. 28, 2015 on Pith
  15. [41] Y. Wan, C. Yao, M. Song, and L. Zhang, ``Non-stationary online convex optimization with arbitrary delays,'' in Proc. Int. Conf. Mach. Learn., 2024, pp. 49\,991--50\,011 on Pith
  16. [42] P. Wu, H. Huang, and Z. Liu, ``Online sequential decision-making with unknown delays,'' in Proc. ACM Web Conf., 2024, pp. 4028--4036 on Pith
  17. [43] A. Ryabchenko, I. Attias, and D. M. Roy, ``Capacity-constrained online learning with delays: Scheduling frameworks and regret trade-offs,'' in Proc. Conf. Learn. Theory, 2025, pp. 4973--5014 on Pith
  18. [45] A. N. Angelopoulos, M. I. Jordan, and R. J. Tibshirani, ``Gradient equilibrium in online learning: Theory and applications,'' J. Mach. Learn. Res., vol. 26, no. 305, pp. 1--68, 2025 on Pith
  19. [46] M. Zecchin, U. K. Ganesan, G. Durisi, P. Popovski, and O. Simeone, ``Prediction-powered communication with distortion guarantees,'' IEEE J. Sel. Areas Inf. Theory, vol. 7, pp. 33--45, 2026 on Pith
  20. [47] V. Elvira and L. Martino, Advances in Importance Sampling. 1em plus 0.5em minus 0.4em John Wiley & Sons, Ltd, 2021, pp. 1--14 on Pith
  21. [48] A. N. Angelopoulos and S. Bates, ``Conformal prediction: A gentle introduction,'' Found. Trends Mach. Learn., vol. 16, no. 4, pp. 494--591, 2023 on Pith
  22. [49] J. Deng, J. Krause, A. C. Berg, and L. Fei-Fei, ``Hedging your bets: Optimizing accuracy-specificity trade-offs in large scale visual recognition,'' in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2012, pp. 3450--3457 on Pith
  23. [50] D.-P. Fan, G.-P. Ji, T. Zhou, G. Chen, H. Fu, J. Shen, and L. Shao, `` PraNet : Parallel reverse attention network for polyp segmentation,'' in Proc. Int. Conf. Med. Image Comput. Comput.-Assist. Intervent., 2020, pp. 263--273 on Pith
  24. [51] V. Karpukhin, B. Oguz, S. Min, P. Lewis, L. Wu, S. Edunov, D. Chen, and W.-t. Yih, ``Dense passage retrieval for open-domain question answering,'' in Proc. Conf. Empir. Methods Nat. Lang. Process., 2020, pp. 6769--6781 on Pith
  25. [52] P. Rajpurkar, J. Zhang, K. Lopyrev, and P. Liang, `` SQuAD : 100,000+ questions for machine comprehension of text,'' in Proc. Conf. Empir. Methods Nat. Lang. Process., 2016, pp. 2383--2392 on Pith
  26. [53] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, `` ImageNet : A large-scale hierarchical image database,'' in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2009, pp. 248--255 on Pith
  27. [54] C. Fellbaum, WordNet: An electronic lexical database. 1em plus 0.5em minus 0.4em MIT press, 1998 on Pith
  28. [55] K. He, X. Zhang, S. Ren, and J. Sun, ``Deep residual learning for image recognition,'' in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2016, pp. 770--778 on Pith
  29. [56] F. Yao, J. Wu, G. Venkataramani, and S. Subramaniam, `` TS-BatPro : Improving energy efficiency in data centers by leveraging temporal–spatial batching,'' IEEE Trans. Cogn. Commun. Netw., vol. 3, no. 1, pp. 236--249, 2019 on Pith
  30. [57] W. Chen, R. F. da Silva, E. Deelman, and T. Fahringer, ``Dynamic and fault-tolerant clustering for scientific workflows,'' IEEE Trans. Cloud Comput., vol. 4, no. 1, pp. 49--62, 2016 on Pith
  31. [58] L. Kleinrock, Queueing Systems, Volume I : Theory . 1em plus 0.5em minus 0.4em New York: Wiley, 1975 on Pith
  32. [59] M. Kalman and B. Girod, ``Modeling the delays of successively-transmitted Internet packets,'' in Proc. IEEE Int. Conf. Multimedia and Expo (ICME), 2004, pp. 2015--2018 on Pith
  33. [60] A. Severinson, E. Rosnes, S. El Rouayheb, and A. G. i Amat, `` DSAG : A mixed synchronous-asynchronous iterative method for straggler-resilient learning,'' IEEE Trans. Commun. , vol. 71, no. 2, pp. 808--822, 2022 on Pith
  34. [61] Y. Liang and B. Girod, ``Network-adaptive low-latency video communication over best-effort networks,'' IEEE Trans. Circuits Syst. Video Technol. , vol. 16, no. 1, pp. 72--81, 2006 on Pith
  35. [62] S. I. Ktena, A. Tejani, L. Theis, P. K. Myana, D. Dilipkumar, F. Husz \'a r, S. Yoo, and W. Shi, ``Addressing delayed feedback for continuous training with neural networks in CTR prediction,'' in Proc. ACM Conf. Recomm. Syst., 2019, pp. 187--195 on Pith
  36. [63] P. A. W. Lewis, E. McKenzie, and D. K. Hugus, ``Gamma processes,'' Communications in Statistics. Stochastic Models, vol. 5, no. 1, pp. 1--30, 1989 on Pith
  37. [64] Capacity-Constrained Online Convex Optimization with Delayed Feedback , author=. arXiv preprint arXiv:2606.11711 , year= on PithPith's reviewarXiv
  38. [65] Wang, Kaizheng and Cuzzolin, Fabio and Shariatmadar, Keivan and Moens, David and Hallez, Hans , journal=IEEE_J_PAMI, title=. 2026 , volume= on Pith
  39. [66] Constrained online convex optimization with feedback delays , author= on Pith
  40. [67] Conformal prediction: A gentle introduction , author=. Found. Trends Mach. Learn. , volume= on Pith
  41. [68] Adaptive Conformal Inference Under Delayed Feedback: Coverage Guarantees and a Delay-to-Memory Diagnostic , author=. arXiv preprint arXiv:2609.07251 , year= on PithPith's reviewarXiv
  42. [69] Online conformal prediction with decaying step sizes , author=. Proc. Int. Conf. Mach. Learn. , pages= on Pith
  43. [70] Adaptive conformal inference under distribution shift , author=. Proc. Adv. Neural Inf. Process. Syst. , volume= on Pith
  44. [71] Mirror online conformal prediction with intermittent feedback , author= on Pith
  45. [72] Conformalized interactive imitation learning: Handling expert shift and intermittent feedback , author=. Proc. Int. Conf. Learn. Represent. , pages= on Pith
  46. [73] Elvira, Víctor and Martino, Luca , publisher =. Advances in Importance Sampling , booktitle = on Pith
  47. [74] Zecchin, Matteo and Ganesan, Unnikrishnan Kunnath and Durisi, Giuseppe and Popovski, Petar and Simeone, Osvaldo , journal=. Prediction-Powered Communication With Distortion Guarantees , year= on Pith
  48. [75] Convex optimization , author=. 2004 , publisher= on Pith
  49. [76] Order restricted statistical inference , author=. Journal of Applied Econometrics , volume = on Pith
  50. [77] Achieving Risk Control in Online Learning Settings , author=. Trans. Mach. Learn. Res. , year= on Pith
  51. [78] Localized Adaptive Risk Control , author=. Proc. Adv. Neural Inf. Process. Syst. , pages = on Pith
  52. [79] Conformal Risk Control , author=. Proc. Int. Conf. Learn. Represent. , pages = on Pith
  53. [80] Exploiting Curvature in Online Convex Optimization with Delayed Feedback , author=. Proc. Int. Conf. Mach. Learn. , pages= on Pith
  54. [81] Wang, Bowen and Zecchin, Matteo and Simeone, Osvaldo , journal=IEEE_J_JSAC, title=. 2026 , volume= on Pith
  55. [82] Fan, Deng-Ping and Ji, Ge-Peng and Zhou, Tao and Chen, Geng and Fu, Huazhu and Shen, Jianbing and Shao, Ling , booktitle= on Pith
  56. [83] Theoretical foundations of conformal prediction , author=. arXiv preprint arXiv:2411.11824 , year= on PithPith's reviewarXiv
  57. [84] Application of conformal prediction interval estimations to market makers’ net positions , author=. Proc. Symp. Conformal Probabilistic Prediction Appl. , pages= on Pith
  58. [85] Conformal decision theory: Safe autonomous decisions from imperfect predictions , author=. Proc. IEEE Int. Conf. Robot. Automat. , pages= on Pith
  59. [86] Simeone, Osvaldo and Park, Sangwoo and Zecchin, Matteo , journal=IEEE_M_COM, title=. 2026 , volume= on Pith
  60. [87] Simeone, Osvaldo and Romano, Yaniv , journal=. Uncertainty Quantification and Data Efficiency in. 2026 , volume= on Pith
  61. [88] Wiesmayr, Reinhard and Maggi, Lorenzo and Cammerer, Sebastian and Hoydis, Jakob and Aoudia, Fay. arXiv preprint arXiv:2510.05784 , year= on PitharXiv
  62. [89] Cifra, Christina L and Sittig, Dean F and Singh, Hardeep , title =. 2021 , journal = on Pith
  63. [90] Calibrated multi-level quantile forecasting , author=. arXiv preprint arXiv:2512.23671 , year= on PitharXiv
  64. [91] Adaptive Conformal Inference for Multi-Step Ahead Time-Series Forecasting Online , author=. Proc. Symp. Conformal Probabilistic Prediction Appl. , pages= on Pith
  65. [92] Slow learners are fast , author=. arXiv preprint arXiv:0911.0491 , year= on PithPith's reviewarXiv
  66. [93] Online learning with adversarial delays , author=. Proc. Adv. Neural Inf. Process. Syst. , volume= on Pith
  67. [94] Non-stationary online convex optimization with arbitrary delays , author=. Proc. Int. Conf. Mach. Learn. , pages= on Pith
  68. [95] Online sequential decision-making with unknown delays , author=. Proc. ACM Web Conf. , pages= on Pith
  69. [96] Capacity-Constrained Online Learning with Delays: Scheduling Frameworks and Regret Trade-offs , author=. Proc. Conf. Learn. Theory , pages= on Pith
  70. [97] Vershynin, Roman , year=. High-Dimensional Probability: An Introduction with Applications in Data Science , publisher= on Pith
  71. [98] Probability: theory and examples , author=. 2019 , publisher= on Pith
  72. [99] Xu, Chen and Xie, Yao , journal=IEEE_J_PAMI, title=. 2023 , volume= on Pith
  73. [100] Einbinder, Bat-Sheva and Ringel, Liran and Romano, Yaniv , journal=IEEE_J_PAMI, title=. 2025 , volume= on Pith
  74. [101] Park, Sangwoo and Cohen, Kfir M. and Simeone, Osvaldo , journal=IEEE_J_PAMI, title=. 2024 , volume= on Pith
  75. [102] Zhu, Meiyi and Zecchin, Matteo and Park, Sangwoo and Guo, Caili and Feng, Chunyan and Popovski, Petar and Simeone, Osvaldo , journal=IEEE_J_SP, title=. 2025 , volume= on Pith
  76. [103] Binucci, Francesco and Simeone, Osvaldo and Banelli, Paolo , journal=IEEE_J_SP, title=. 2025 , volume= on Pith
  77. [104] Automatically Adaptive Conformal Risk Control , author=. Proc. Int. Conf. Artif. Intell. Statist. , year= on Pith
  78. [105] A modern introduction to online learning , author=. arXiv preprint arXiv:1912.13213 , year= on PithPith's reviewarXiv
  79. [106] Online learning and online convex optimization , author=. Found. Trends Mach. Learn. , volume= on Pith
  80. [107] Online convex programming and generalized infinitesimal gradient ascent , author=. Proc. Int. Conf. Mach. Learn. , pages= on Pith

The first 100 of 125 references are listed here. All of them are in this paper's citation record.

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment