Pith. sign in

REVIEW 3 major objections 5 minor 51 references

EulerLoRA: Rank-Driven Jump Dynamics for Calibrated Parameter-Efficient Fine-Tuning

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Rank-sampled LoRA matches a 16-model ensemble at 69% lower cost.

desk verdict EulerLoRA is a genuinely new mean-preserving stochastic LoRA construction with clean math and honest caveats; the full-block unbiasedness is only projection-level, and stochastic inference gains are narrower than the abstract implies, but the deterministic trained model alone is competitive and the paper deserves a serious referee. read the letter →

arxiv 2608.01142 v2 pith:5M2OHQTV submitted 2026-08-02 cs.LG

Srinivas Anumasa , Dianbo Liu This is my paper · ORCID
classification cs.LG
keywords parameter-efficientfine-tuningLoRApredictiveuncertaintyensemblemethodsrank-drivendynamicsmodelcalibrationout-of-distributiondetectionvisiontransformers
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 tries to establish that predictive diversity—the usual benefit of training many separate models—can be obtained from a single shared low-rank adapter by adding structured randomness along its rank-one components. The method, EulerLoRA, evolves an auxiliary state through a few internal Euler steps, sampling a subset of rank directions at each step and rescaling their coefficients so that every direction contributes with expected coefficient one; the average over trajectories is exactly the deterministic LoRA transformation. If this holds, then calibration, negative log-likelihood, and out-of-distribution detection can be improved without training one adapter per ensemble member. The paper reports that two rank-20 adapters (about 3 million trainable parameters) match or beat a 16-adapter rank-8 LoRA-Ensemble (about 10 million parameters) on CIFAR-10, CIFAR-100, HAM10000, and SVHN out-of-distribution detection.

What carries the argument

The central object is the compensated rank-driven jump dynamics at the output projection. An auxiliary state $Y_{\ell,s}$ evolves for $T$ internal Euler steps with $\Delta t=1/T$; each step samples a rank configuration $(K_{\ell,s},S_{\ell,s})$ and applies the coefficient $M_{\ell,s,k}=r/K_{\ell,s}$ to selected rank-one directions. Writing $M=1+(M-1)$ separates the predictable LoRA increment (accumulated once over the unit interval via $\Delta t$ scaling) from centered zero-mean noise (accumulated with $\sqrt{\Delta t}$ scaling), which is what makes the terminal expectation exactly LoRA while individual trajectories fluctuate. In deterministic inference the sampled subset is all $r$ components, so $M=1$ and the stochastic term vanishes; in stochastic inference, four Monte Carlo samples per adapter are averaged. The construction is applied to the output projection with $T=2$; query, key, and value projections use a single expectation-preserving rank-sampled update per forward pass.

What would settle it

Measure the gap between the stochastic block's average output and the deterministic block's output on a real attention block: compute $\|\mathbb{E}[\mathrm{softmax}(Y_{\ell,T})]-\mathrm{softmax}(\mathbb{E}[Y_{\ell,T}])\|$ with a large number of Monte Carlo samples; if this gap stays large while trajectory diversity is high and stochastic inference never improves calibration or OOD detection over deterministic inference across ranks and datasets, the claim that shared-adapter rank sampling produces useful predictive diversity would be refuted.

Watch

Extended reading notes

Core claim

At the output projection of a transformer block, EulerLoRA replaces the deterministic LoRA correction $B_\ell A_\ell C_\ell$ with a finite-step stochastic process. At each of $T$ internal steps it samples an active rank $K_{\ell,s}$ and a uniformly chosen subset of rank-one directions, giving component coefficients $M_{\ell,s,k}=\frac{r}{K_{\ell,s}}\mathbf{1}\{k\in S_{\ell,s}\}$ with conditional expectation one. The update separates the deterministic LoRA term (scaled by $\Delta t$) from the centered fluctuation $M_{\ell,s,k}-1$ (scaled by $\sqrt{\Delta t}$), so the expected terminal state satisfies $\mathbb{E}[Y_{\ell,T}\mid C_\ell]=h_\ell+W^0_{O,\ell}C_\ell+B_\ell A_\ell C_\ell$, exactly the deterministic LoRA block, independent of $T$. Trajectories differ around that mean, and averaging softmax probabilities over them produces an implicit ensemble. The paper's empirical claim is that this shared-adapter diversity is useful: two rank-20 adapters outperform or match a much larger 16-adapter LoRA-Ensemble in accuracy, calibration, NLL, Brier score, and OOD detection, with about 69% fewer trainable adapter parameters.

Load-bearing premise

The practical gains rest on the assumption that the nonlinear operations inside self-attention, especially softmax, do not destroy the diversity or bias the average of the rank-sampled trajectories, so the projection-level expectation guarantee still yields useful predictive diversity at the block output.

Editorial extensions

If this is right

  • The number of Monte Carlo predictions is decoupled from the number of learned adapters: four samples per adapter yield the reported uncertainty estimates, and more samples can be added without training anything new.
  • One trained EulerLoRA checkpoint supports two inference modes: deterministic inference (all rank components active) gives the best accuracy, NLL, and Brier score, while stochastic inference gives the best calibration on HAM10000 and the best OOD FPR@95TPR on SVHN.
  • Two rank-20 adapters (about 2.95 million adapter parameters) match or beat a 16-adapter rank-8 LoRA-Ensemble (about 9.44 million adapter parameters), a reduction of about 69% in trainable adapter parameters.
  • On CIFAR-10, EulerLoRA with two adapters reaches 96.10% accuracy and 0.116 NLL in deterministic mode, improving on the reproduced two-adapter LoRA-Ensemble (95.43%, 0.138) and the published 16-adapter ensemble (95.9%, 0.128).
  • On SVHN OOD detection, stochastic EulerLoRA lowers FPR@95TPR to 37.42%, a 7.76-percentage-point improvement over the reproduced two-adapter LoRA-Ensemble.

Reading between the lines

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

  • The mean-preservation theorem is proved only at the linear projection level; because self-attention applies softmax, the full block output is not guaranteed to be an unbiased estimate of the deterministic block. A direct extension would be to quantify the bias $\mathbb{E}[\mathrm{softmax}(Y_{\ell,T})]-\mathrm{softmax}(\mathbb{E}[Y_{\ell,T}])$ and test whether it shrinks with more internal steps or
  • EulerLoRA's rank-sampling schedule is a free design choice: the paper fixes $r=20$, $K_{\min}=10$, and uniform $K$. Varying the minimum active rank or making the coefficient distribution non-uniform would change trajectory diversity, so the reported parameter-efficiency gain might extend to even smaller adapters or collapse if trajectories become correlated.
  • Because the expected terminal state is exactly the deterministic LoRA block, EulerLoRA can be layered on top of existing LoRA training pipelines as a drop-in stochastic head: the same checkpoint can serve as a deterministic predictor and, at inference time, as an uncertainty estimator, which suggests applications to active learning or selective prediction where calibrated confidence is needed.
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 / 5 minor

Summary. The paper introduces EulerLoRA, a stochastic extension of LoRA. At the output projection of each attention block, an auxiliary state is evolved over T internal Euler steps; at each step a random subset of rank-one LoRA components is activated with weights r/K, and the centered fluctuations are added with stochastic scaling. The paper proves (Eq. 17) that, conditional on the input to the output projection, the expected terminal state equals the deterministic LoRA block. During stochastic inference, multiple forward passes are averaged to form an implicit ensemble; setting all rank components active gives deterministic inference. Experiments on ViT-B/32 compare one- and two-adapter rank-20 configurations against LoRA-Ensemble baselines on CIFAR-10, CIFAR-100, HAM10000, and SVHN OOD detection, reporting accuracy, macro-F1, ECE, NLL, Brier score, and OOD metrics. The headline claim is that two rank-20 adapters (~3M trainable parameters) match or beat a rank-8, 16-adapter LoRA-Ensemble (~10M parameters), a reduction of about 69%.

Significance. If the claims were fully established, the main contribution would be a simple and parameter-efficient way to obtain predictive diversity from shared LoRA adapters, with a clean projection-level expectation guarantee and a deterministic inference mode. The paper is transparent in several respects: it reports five-seed means and standard deviations, reproduces rank-20 baselines, gives detailed hyperparameters and training settings, and explicitly states in the supplementary that the expectation guarantee does not extend through the attention softmax. However, the theoretical guarantee is narrower than the abstract suggests, and the stochastic-inference advantage is not uniform across benchmarks. The value of the paper after revision would be as an honest demonstration that stochastic training of LoRA can improve deterministic performance and sometimes improve calibration and OOD detection, rather than as a general unbiased-ensemble method.

major comments (3)
  1. [Compensated Rank-Driven Jump Dynamics, Eq. (17); supplementary 'Random Active-Rank Sampling at Q, K, and V'] The central theoretical claim is proven only for the output projection conditional on C_l. Eq. (24) samples rank configurations at the Q, K, and V projections in every stochastic forward pass, and those stochastic inputs pass through the attention softmax before C_l is formed. Since softmax is nonlinear, the expectation of the complete stochastic attention block is not the deterministic attention block; the supplementary explicitly concedes this. The abstract and contributions, however, state that EulerLoRA preserves 'the deterministic LoRA transformation in expectation' without this qualification. This is a load-bearing overclaim: the Monte Carlo ensemble at inference is not guaranteed to be centered on the deterministic LoRA prediction, and Tables 1 and 4 show that deterministic inference often beats stochastic inference on proper-scoring metrics. The authors should either qualify the theoretical claims throughout or add an empirical analysis quantifying the full-block bias, for example by comparing the mean stochastic attention output with the deterministic attention output over the validation set.
  2. [Experiments, Tables 1, 3, and 4] The empirical case for stochastic inference as a source of 'useful predictive diversity' is mixed. On CIFAR-10 with two adapters, the stochastic mode is worse than the deterministic mode on every proper-scoring metric (NLL 0.126 vs 0.116, ECE 0.016 vs 0.004, Brier 0.064 vs 0.059). On CIFAR-100, the stochastic mode has substantially higher ECE (0.058 vs 0.013) and NLL (0.589 vs 0.559). The clear stochastic-inference wins are HAM10000 calibration/NLL and SVHN FPR@95TPR. The paper's own 'complementary benefits' conclusion is defensible, but the abstract's general claim that 'useful predictive diversity can be obtained' needs to be scoped, ideally with statistical significance tests or effect sizes over the five seeds and with a discussion of when stochastic inference is expected to help.
  3. [Experiments, Baselines and Tables 1, 3, 4] The headline comparison with the 16-adapter LoRA-Ensemble uses numbers taken from the original LoRA-Ensemble paper rather than re-run under the same protocol and seeds. The reproduced controls are rank-20 LoRA and rank-20 two-adapter LoRA-Ensemble. Given the central parameter-efficiency claim, a controlled reimplementation of the rank-8/16-adapter baseline with the same five seeds, training script, and evaluation pipeline would substantially strengthen the claim; as it stands, the cross-paper comparison may conflate implementation differences with method differences.
minor comments (5)
  1. [Figure 1] The sentence 'Figure 1 gives an overview our proposed method' should read 'gives an overview of our proposed method'.
  2. [Supplementary Material, Evaluation Metrics] The reproducibility summary says 'Compute the metrics in Section' without a section number or name; this should be completed.
  3. [Experiments, Tables 1 and 4] No statistical significance testing is reported; given five seeds and small standard deviations, statements comparing close values such as CIFAR-100 accuracy 82.6 +/- 0.2 vs 82.6 +/- 0.1 should avoid over-interpreting small gaps.
  4. [Overall analysis] The paper does not report diversity diagnostics such as average pairwise disagreement or variance of the Monte Carlo predictions, even though predictive diversity is the core mechanism; reporting one such measure would directly support the narrative.
  5. [Training protocol] The training procedure uses four stochastic samples per adapter, but the compute overhead of training is not reported, so the parameter-efficiency comparison alone does not fully characterize the total cost relative to LoRA-Ensemble.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: EulerLoRA's mean-preserving guarantee is an explicit design identity, and the empirical comparisons use external or reproduced controls.

full rationale

The derivation of Eq. (17) is self-contained. The sampling rule M_{ℓ,s,k} = (r/K_{ℓ,s})1{k∈S} is chosen so that E[M|K] = 1 (Eq. 9), and the compensated dynamics in Eq. (15) explicitly include the deterministic LoRA increment with Euler scaling while the centered fluctuation (M−1) has zero mean. Summing over T steps therefore yields Eq. (17) by elementary probability, not by importing the conclusion. The paper transparently labels this as a construction ('we avoid this by separating the deterministic LoRA transformation from the centered stochastic fluctuations'), so the mean-preservation claim is a definitional theorem rather than a fitted prediction. No parameter is fitted to the benchmark metrics: T, σ, K_min, and S are fixed hyperparameters, and the comparison against LoRA-Ensemble uses published or explicitly reproduced controls. The supplementary's concession that projection-level unbiasedness does not extend through the attention softmax is a limitation on the end-to-end unbiasedness claim, not a circular step, since the paper does not define the method in terms of the benchmarks. There is no self-citation chain supplying the central result.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method introduces mathematical constructs (auxiliary state, rank configurations) but no new physical or model-level entities. The free parameters are hand-chosen hyperparameters, not fitted to the benchmark results. The axioms are standard background facts plus one design choice specific to this paper.

free parameters (6)
  • Maximum LoRA rank r = 20
    Set to 20 for all controlled experiments; the central parameter-efficiency comparison (two rank-20 adapters vs sixteen rank-8 adapters) depends on this choice.
  • Minimum active rank Kmin = 10
    The active rank is sampled uniformly from {10,...,20}; the lower bound is chosen by hand, and no sensitivity analysis is reported.
  • Stochastic scale sigma = 1.0
    Controls the magnitude of the centered rank fluctuations; chosen without a reported sweep.
  • Internal steps T = 2
    Number of compensated dynamics steps at the output projection; the expectation-preservation property holds for any T, but the empirical results use T=2.
  • Monte Carlo samples per adapter S = 4
    Number of stochastic forward passes averaged at inference; no analysis of the effect of S is provided.
  • Number of adapters A = 2 (CIFAR-100, HAM10000) or 1-2 (CIFAR-10)
    The headline comparison uses two adapters; the claim of 69% fewer parameters is relative to A=2.
assumptions (4)
  • standard math The LoRA correction can be decomposed into r rank-one components, each acting as an independent learned direction.
    Eq. (5) uses the linear-algebra fact that a rank-r matrix product decomposes into a sum of rank-one terms.
  • standard math Sampling K components uniformly and scaling by r/K keeps each component unchanged in expectation.
    Eqs. (8)-(9); a straightforward probability calculation.
  • domain assumption The residual update can be interpreted as a forward-Euler discretization, so inserting finite internal steps along the additive output path is a valid construction.
    Motivational framing from neural ODE literature; the construction is defined by Eq. (15) regardless of the ODE interpretation.
  • ad hoc to paper Only the output projection receives the T-step compensated dynamics; Q, K, V projections use a single rank-sampled update.
    Stated in the supplementary; this design choice is not derived and is not justified beyond an implementation preference.

how reviews work

0 comments
Cite this review

Pith. "Pith review of EulerLoRA: Rank-Driven Jump Dynamics for Calibrated Parameter-Efficient Fine-Tuning." pith.science (2026). https://pith.science/paper/5M2OHQTV

@misc{pith2026260801142,
  author       = {Pith},
  title        = {Pith review of: EulerLoRA: Rank-Driven Jump Dynamics for Calibrated Parameter-Efficient Fine-Tuning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5M2OHQTV}},
  note         = {Machine review of arXiv:2608.01142}
}
read the original abstract

Low-Rank Adaptation (LoRA) enables parameter-efficient fine-tuning, but standard LoRA produces a single deterministic model and does not directly support predictive uncertainty estimation. We introduce EulerLoRA, a stochastic extension of LoRA that generates multiple predictive trajectories by sampling structured variations along the rank-one components of shared low-rank adapters, while preserving the deterministic LoRA transformation in expectation. We evaluate EulerLoRA with vision transformers on CIFAR-10, CIFAR-100, and HAM10000, together with out-of-distribution detection on SVHN. Across these benchmarks, EulerLoRA achieves comparable or improved performance relative to strong LoRA-Ensemble baselines. Using two rank-20 adapters, EulerLoRA requires approximately 3 million trainable adapter parameters, compared with about 10 million for a rank-8, 16-adapter LoRA-Ensemble, corresponding to roughly 69% fewer trainable parameters. These results show that useful predictive diversity can be obtained from a small number of shared adapters.

Figures

Figures reproduced from arXiv: 2608.01142 by the authors.

Figure 1
Figure 1. Overview of the proposed compensated rank-driven dynamics for LoRA. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 20 canonical work pages

  1. [1]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  2. [2]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  3. [3]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  4. [4]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  5. [5]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    doi:10.1016/s0020-7373(84)80003-6 1984

  6. [6]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

  7. [7]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  8. [8]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

Show all 51 references
  1. [9]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  2. [10]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  3. [11]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

  4. [12]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. Iclr , year=

  5. [13]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

    Deep Residual Learning for Image Recognition , author =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages =

  6. [14]

    Science Advances , volume=

    Spikingjelly: An open-source machine learning infrastructure platform for spike-based intelligence , author=. Science Advances , volume=. 2023 , publisher=

    2023

  7. [15]

    Advances in Neural Information Processing Systems , volume =

    Neural Ordinary Differential Equations , author =. Advances in Neural Information Processing Systems , volume =

  8. [16]

    arXiv preprint arXiv:1812.00174 , year =

    Stochastic Training of Residual Networks: A Differential Equation Viewpoint , author =. arXiv preprint arXiv:1812.00174 , year =

    arXiv

  9. [17]

    How Does Noise Help Robustness? Explanation and Exploration under the Neural

    Liu, Xuanqing and Xiao, Tesi and Si, Si and Cao, Qin and Kumar, Sanjiv and Hsieh, Cho-Jui , booktitle =. How Does Noise Help Robustness? Explanation and Exploration under the Neural

  10. [18]

    Advances in Neural Information Processing Systems , volume =

    Neural Jump Stochastic Differential Equations , author =. Advances in Neural Information Processing Systems , volume =

  11. [19]

    Proceedings of the 33rd International Conference on Machine Learning , series =

    Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , author =. Proceedings of the 33rd International Conference on Machine Learning , series =

  12. [20]

    Advances in Neural Information Processing Systems , volume =

    Simple and Scalable Predictive Uncertainty Estimation Using Deep Ensembles , author =. Advances in Neural Information Processing Systems , volume =

  13. [21]

    International Conference on Learning Representations , year =

    Bayesian Low-Rank Adaptation for Large Language Models , author =. International Conference on Learning Representations , year =

  14. [22]

    and Wang, Hao , booktitle =

    Wang, Yibin and Shi, Haizhou and Han, Ligong and Metaxas, Dimitris N. and Wang, Hao , booktitle =

  15. [23]

    arXiv preprint arXiv:2410.03941 , year =

    Kasymov, Artur and Sendera, Marcin and Stypu. arXiv preprint arXiv:2410.03941 , year =

    arXiv

  16. [24]

    Proceedings of NAACL-HLT , year =

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , author =. Proceedings of NAACL-HLT , year =

  17. [25]

    Advances in Neural Information Processing Systems , volume =

    Language Models are Few-Shot Learners , author =. Advances in Neural Information Processing Systems , volume =

  18. [26]

    Journal of Machine Learning Research , volume =

    Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer , author =. Journal of Machine Learning Research , volume =

  19. [27]

    Proceedings of the International Conference on Machine Learning , year =

    Learning Transferable Visual Models from Natural Language Supervision , author =. Proceedings of the International Conference on Machine Learning , year =

  20. [28]

    Zhang, Qingru and Chen, Minshuo and Bukharin, Alexander and Karampatziakis, Nikos and He, Pengcheng and Cheng, Yu and Chen, Weizhu and Zhao, Tuo , booktitle =

  21. [29]

    Valipour, Mojtaba and Rezagholizadeh, Mehdi and Kobyzev, Ivan and Ghodsi, Ali , booktitle =

  22. [30]

    Inverse Problems , volume =

    Stable Architectures for Deep Neural Networks , author =. Inverse Problems , volume =

  23. [31]

    Advances in Neural Information Processing Systems , volume =

    Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift , author =. Advances in Neural Information Processing Systems , volume =

  24. [32]

    International conference on machine learning , pages=

    Beyond finite layer neural networks: Bridging deep architectures and numerical differential equations , author=. International conference on machine learning , pages=. 2018 , organization=

    2018

  25. [33]

    Liu, Shih-Yang and Wang, Chien-Yi and Yin, Hongxu and Molchanov, Pavlo and Wang, Yu-Chiang Frank and Cheng, Kwang-Ting and Chen, Min-Hung , booktitle =

  26. [34]

    International Conference on Machine Learning , author =

    Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , year =. International Conference on Machine Learning , author =

  27. [35]

    Advances in Neural Information Processing Systems , author =

    Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles , year =. Advances in Neural Information Processing Systems , author =

  28. [36]

    International Conference on Learning Representations , author =

    Training independent subnetworks for robust prediction , year =. International Conference on Learning Representations , author =

  29. [37]

    arXiv preprint arXiv:2107.08924 , year=

    Epistemic neural networks , author=. arXiv preprint arXiv:2107.08924 , year=

    arXiv

  30. [38]

    Advances in Neural Information Processing Systems , volume=

    Epistemic neural networks , author=. Advances in Neural Information Processing Systems , volume=

  31. [39]

    Advances in Neural Information Processing Systems , volume=

    Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness , author=. Advances in Neural Information Processing Systems , volume=

  32. [40]

    International Conference on Machine Learning , pages=

    The Lipschitz Constant of Self-Attention , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  33. [41]

    Mitigating Transformer Overconfidence via

    Ye, Wenqian and Ma, Yunsheng and Cao, Xu and Tang, Kun , booktitle =. Mitigating Transformer Overconfidence via

  34. [42]

    , title =

    Brier, Glenn W. , title =. Monthly Weather Review , volume =. 1950 , doi =

    1950

  35. [43]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Cui, Yin and Jia, Menglin and Lin, Tsung-Yi and Song, Yang and Belongie, Serge , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

  36. [44]

    International Conference on Learning Representations , year =

    Dosovitskiy, Alexey and Beyer, Lucas and Kolesnikov, Alexander and Weissenborn, Dirk and Zhai, Xiaohua and Unterthiner, Thomas and Dehghani, Mostafa and Minderer, Matthias and Heigold, Georg and Gelly, Sylvain and Uszkoreit, Jakob and Houlsby, Neil , title =. International Con...

  37. [45]

    , title =

    Guo, Chuan and Pleiss, Geoff and Sun, Yu and Weinberger, Kilian Q. , title =. Proceedings of the 34th International Conference on Machine Learning , series =

  38. [46]

    Transactions on Machine Learning Research , year =

    M. Transactions on Machine Learning Research , year =

  39. [47]

    International Conference on Learning Representations , year =

    Hendrycks, Dan and Gimpel, Kevin , title =. International Conference on Learning Representations , year =

  40. [48]

    Krizhevsky, Alex , title =

  41. [49]

    International Conference on Learning Representations , year =

    Loshchilov, Ilya and Hutter, Frank , title =. International Conference on Learning Representations , year =

  42. [50]

    , title =

    Netzer, Yuval and Wang, Tao and Coates, Adam and Bissacco, Alessandro and Wu, Bo and Ng, Andrew Y. , title =. NIPS Workshop on Deep Learning and Unsupervised Feature Learning , year =

  43. [51]

    Scientific data , volume=

    The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions , author=. Scientific data , volume=. 2018 , publisher=

    2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.