ITP: Instance-Aware Test Pruning for Out-of-Distribution Detection

Haonan Xu; Yang Yang

REVIEW 3 major objections 6 minor 47 references

ITP: Instance-Aware Test Pruning for Out-of-Distribution Detection

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

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

Pith's one-line read Pruning overconfident weights at test time improves OOD detection.

desk verdict Solid, practical OOD pruning method with honest ablations, but the universal improvement claim fails on Places365 and the SOTA claim leans on appendix-only baselines. read the letter →

arxiv 2412.12566 v1 pith:UETDOFAB submitted 2024-12-17 cs.CV

Haonan Xu , Yang Yang This is my paper · ORCID

classification cs.CV

keywords out-of-distributiondetectiontest-timepruningweightsparsificationZ-scoretestparametercontributiondistributionpost-hocmethodoverconfidenceenergyscore

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

The paper claims that a post-hoc, training-free pruning scheme can substantially improve out-of-distribution detection in pre-trained image classifiers. Its diagnosis is that parameters which are important for in-distribution predictions can still behave overconfidently on OOD inputs, inflating the model's ID confidence. To fix this, ITP first removes redundant last-layer weights using class-specific contribution statistics, then, for each test sample, prunes weights whose contribution is an outlier high according to a right-tailed Z-score test. The result is a wider separation between ID and OOD energy scores, reported as large reductions in FPR95 on CIFAR and ImageNet benchmarks. The significance is that better OOD rejection is achieved without retraining or access to OOD data.

What carries the argument

The machinery is the class-specific parameter contribution distribution. For a weight $W_{ij}$ in the last fully connected layer, its contribution to class $j$ for input $x$ is $W_{ij} h_i(x)$; the paper estimates mean $\mu_{ij}$ and $\sigma_{ij}$ over class-$j$ training data, assuming approximate Gaussianity. Coarse pruning (CRP) zeroes weights whose mean contribution falls in the lowest $p$ percentile, while fine-grained test pruning (FTP) zeroes weights whose standardized contribution $(c - \mu)/\sigma$ exceeds $\lambda$, a right-tailed Z-score test. The final OOD score is the negative energy of the logits computed with the element-wise product of both masks, which adaptively suppresses both noise and overconfidence per sample.

What would settle it

Take a trained classifier and its class-specific contribution statistics, then compute the two sides of Equation (14) on a fresh OOD benchmark. If the summed contributions of the pruned parameters are not larger for OOD than for ID inputs, the claimed mechanism for CRP is false; alternatively, show that setting those same parameters to zero raises FPR95 compared with keeping them.

Watch

Extended reading notes

Core claim

The central discovery is that OOD detection can be improved by pruning, at test time, parameters that are overconfident with respect to the in-distribution contribution distribution, not only by pruning parameters that are under-contributing. The paper defines each last-layer weight's contribution to a class output as its multiplicative effect on that class's logit, estimates a per-class Gaussian model of these contributions from ID data, and uses two masks: a fixed coarse mask that drops the lowest-contribution weights and an instance-specific fine mask that drops weights whose contribution exceeds the mean by more than a threshold lambda in standardized units. The pruned logits yield OOD scores that separate better, especially on CIFAR-100, where ITP lowers FPR95 from 68.45 (Energy baseline) to 35.03, and on ImageNet where average FPR95 drops to 22.51.

Load-bearing premise

The load-bearing premise is the inequality in Remark 1: parameters with the lowest average contribution to in-distribution predictions contribute more to the logits of OOD inputs than to those of ID inputs.

Editorial extensions

If this is right

ITP improves OOD detection on top of energy, MSP, ODIN, GradNorm, and MLS scores without retraining.
With only 30 images per class used to estimate distributions, ITP retains nearly full performance while preprocessing overhead drops to a few minutes.
ITP is complementary to activation rectification; ITP combined with ReAct beats strong sparsification baselines on CIFAR-100 and ImageNet.
The method adds negligible test-time cost because pruning is a masked matrix multiplication and requires no backpropagation.

Reading between the lines

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

Because FTP is instance-dependent, it could be reused for selective prediction and calibrated rejection in tasks beyond OOD detection, such as detecting misclassified or ambiguous inputs.
The Gaussian assumption in contribution distributions could be relaxed to heavier-tailed families; the Z-score test would then be replaced by a quantile-based rule, potentially improving robustness to outliers.
It may be possible to extend the same two-stage pruning to earlier layers or to neuron-level dimensions, where redundant and overconfident signals also reside.

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.

Desk Editor's Note

read the letter

Short version: ITP is a genuine incremental contribution—combining a fixed redundancy mask with an instance-specific Z-score mask for test-time pruning—and the empirical work is mostly careful. But the central theoretical justification is an unproven inequality, and the headline numbers hide systematic failures on Places365. The SOTA claim is too strong given KNN and ASH only appear in the appendix.

What's new: DICE and LINe prune weights that matter for ID; ITP adds a second step that zeroes out weights whose per-instance contribution is far above the class-specific mean, i.e. parameters that are overconfident on that particular input. That's a real twist. The contribution formula is clean and the ablation shows both CRP and FTP help, especially on CIFAR-100. The preprocessing overhead is genuinely low and the code is public. The paper also shows the method is compatible with other OOD scores, which is useful.

Soft spots, in order of severity:

1. The theory. Remark 1 rests on Eq. 14, the assertion that pruned redundant parameters contribute more to ID-class logits for OOD than for ID. That's asserted, not proven, and it is load-bearing: if the inequality fails, CRP could remove useful information. The paper says it is 'empirically validated' in Figure 3, but Figure 3 shows two cherry-picked ImageNet OOD sets, not the failure cases.

2. The per-dataset results contradict the universal claim. On CIFAR-10 and CIFAR-100 with Places365 as OOD, ITP is worse than plain Energy (45.75 vs 39.40 FPR95, and 82.39 vs 79.20). The averages in Tables 1-2 are pulled up by big wins on SVHN, Textures, etc. This tells you the assumption behind FTP/CRP doesn't hold for every OOD type. The paper never mentions this.

3. SOTA claim. In the main tables the strongest baselines KNN and ASH are absent; in the appendix ASH-S beats ITP on CIFAR-10 FPR95 (15.05 vs 16.72) and ties on ImageNet. The claim 'state-of-the-art' needs qualification.

4. Hyperparameters p and lambda are chosen on a Gaussian-noise validation set. That's a reasonable model-selection trick, but it's a tuned threshold, not a principled statistical test as the 'right-tailed Z-score test' framing implies.

Who it's for: researchers working on post-hoc OOD detection, especially on sparsification-based methods. They'll find the idea useful and reproducible.

Recommendation: definitely send to peer review. It deserves referees, but the authors should be asked to address the per-dataset breakdown, include KNN/ASH in main tables, and either prove or soften Eq. 14.

Referee Report

3 major / 6 minor

Summary. The paper proposes ITP, a post-hoc OOD detection method that prunes last-layer weights. It first estimates class-specific parameter contribution distributions from ID data (assuming Gaussianity). CRP prunes parameters with lowest average contribution; FTP prunes, per test sample, parameters whose contribution Z-score exceeds a threshold λ. The OOD score is negative energy of the pruned logits. Experiments on CIFAR (DenseNet-101) and ImageNet (ResNet-50) show average FPR95 improvements over baselines, with ablations confirming both components.

Significance. If the claims hold, ITP is a useful addition to the post-hoc OOD detection toolbox: it is simple, training-free, has low preprocessing overhead, and is compatible with existing scores and ReAct. The derivation of the contribution formula (Eq. 5) is correct, and the ablations are informative. However, the central mechanism is justified only by an unproven inequality, and the method does not improve over the energy baseline on Places365 in the CIFAR experiments, so the claim that ITP 'improves OOD detection' needs to be scoped.

major comments (3)

[§3.4, Remark 1, Eq. (14)] The inequality in Eq. (14) is asserted without proof and is the sole justification for the benefit of CRP. It is an empirical claim about OOD inputs that cannot be guaranteed from ID statistics alone. The paper's own per-dataset results (Tables 7 and 8) show that this mechanism fails on Places365: ITP has FPR95 45.75 vs Energy 39.40 on CIFAR-10, and 82.39 vs 79.20 on CIFAR-100. Reporting only averages in Tables 1 and 2 conceals these systematic failures. The authors should either prove Eq. (14) under stated conditions, validate it per OOD dataset, or explicitly scope the method's benefit to OOD types where the inequality holds.
[§3.3, Eqs. (9)-(10)] The 'right-tailed Z-score test' is not a statistical hypothesis test: it is a deterministic threshold on a single observation using estimated moments, with λ selected on a validation set. No p-value, significance level, or error-rate control is provided, and the Gaussianity assumption is only checked visually (Appendix B). This framing overstates the method's statistical grounding; the authors should either replace it with a heuristic threshold description or provide proper statistical justification.
[Appendix E and Table 4] The hyperparameters p and λ are selected on Gaussian-noise images. The paper does not justify that Gaussian noise is a valid proxy for the OOD distributions used in evaluation, and the results are highly sensitive to these choices (Table 4 shows ImageNet FPR95 ranging from 22.51 to 73.19 across the grid). The authors should justify the validation strategy or analyze sensitivity to the choice of validation OOD set, since the reported improvements depend on tuned hyperparameters.

minor comments (6)

[Tables 7 and 8] The per-dataset failures on Places365 are not discussed in the main text; the authors should mention them and provide a possible explanation, as they bear on the scope of the method's effectiveness.
[§3.3, after Eq. (9)] The phrase 'single-sample hypothesis testing task' is misleading; the procedure is better described as a fixed threshold rule on a Z-score.
[Tables 9 and 10] There is a typo in the table titles: 'benckmarks' should be 'benchmarks'.
[Figure 1 and Appendix B] The caption mentions 'pre-ReLU activations' but does not explain why these are used for visualization; please define this in the main text or figure caption.
[Appendix E] Please specify the size of the Gaussian-noise validation images and whether they are preprocessed (resized/cropped) in the same way as ID images.
[§4.3, ablation discussion] The explanation for FTP's smaller improvement on ResNet-50 ('larger feature space ... noise dominates') is speculative; consider providing supporting analysis or tempering the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ITP's OOD scores are computed from pruning masks built on ID training data only, and the test OOD sets are not used in fitting.

full rationale

The claimed derivation chain is self-contained. CRP masks are defined by thresholding the mean contribution mu_ij estimated from class-j training data (Eqs. 6-8); FTP masks are defined by a right-tailed Z-score test whose mean and std are also from ID training data (Eqs. 9-10). The final OOD score is the negative energy of the pruned logits (Eqs. 3 and 12) and is evaluated on held-out OOD datasets that are not used when constructing the masks. Hyperparameters p and lambda are selected on a Gaussian-noise validation set (Appendix E); this is model selection on a proxy OOD distribution, not fitting to the test OOD sets, so the average FPR95/AUROC numbers remain an independent evaluation. The theoretical Remarks are heuristic: Remark 1's inequality (Eq. 14) is an asserted empirical property of redundant parameters, not a consequence of the definitions; Remark 2's inequality (Eq. 15) is essentially a restatement of the pruning rule and is not used to force the experimental outcome. The paper's self-citations (Yang et al. 2021/2023a/2023b/2024) appear only as general related-work references and are not load-bearing. The fact that the central assumption can fail on a particular OOD set (e.g., Places365, where ITP is worse than Energy on CIFAR-10/100 in Tables 7-8) concerns correctness/robustness, not circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method rests on one standard scoring assumption (energy), two empirical assumptions about parameter behavior (Gaussian contributions and Eq. 14), and a validation-set proxy (Gaussian noise) for choosing hyperparameters. No new entities are introduced.

free parameters (2)

p (CRP pruning percentile) = p=10 (CIFAR-10), p=20 (CIFAR-100), p=30 (ImageNet-1k)
Fraction of lowest-mean-contribution parameters pruned in coarse-grained pruning. Selected by grid search on a Gaussian-noise validation set to minimize FPR95 (Appendix E).
lambda (FTP Z-score threshold) = lambda=2.2 (CIFAR-10), lambda=1.6 (CIFAR-100), lambda=1.5 (ImageNet-1k)
Threshold for pruning overconfident parameters via right-tailed Z-score. Tuned jointly with p on the Gaussian-noise validation set.

assumptions (4)

domain assumption Parameter contributions of last-layer weights approximately follow Gaussian distributions per class
Stated in Section 3.2 and Appendix B based on visual inspection of histograms (Figure 5); used to justify the Z-score thresholding in FTP.
domain assumption Redundant parameters (lowest average contribution on ID) contribute more to ID-class logits for OOD inputs than for ID inputs (Eq. 14)
Assumed in Remark 1, Section 3.4, with no proof; it is the mechanism by which CRP is claimed to widen the ID/OOD score gap.
ad hoc to paper Gaussian-noise images are a valid proxy for OOD data for hyperparameter selection
Appendix E uses N(0,1) pixel noise as the validation set to pick p and lambda; this assumes such noise is representative of the OOD test distributions.
standard math Energy score is a reliable OOD scoring function
Adopted from Liu et al. 2020, Section 3.1; used as the default OOD score throughout.

how reviews work

0 comments

Cite this review

Pith. "Pith review of ITP: Instance-Aware Test Pruning for Out-of-Distribution Detection." pith.science (2026). https://pith.science/paper/UETDOFAB

@misc{pith2026241212566,
  author       = {Pith},
  title        = {Pith review of: ITP: Instance-Aware Test Pruning for Out-of-Distribution Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UETDOFAB}},
  note         = {Machine review of arXiv:2412.12566}
}

read the original abstract

Out-of-distribution (OOD) detection is crucial for ensuring the reliable deployment of deep models in real-world scenarios. Recently, from the perspective of over-parameterization, a series of methods leveraging weight sparsification techniques have shown promising performance. These methods typically focus on selecting important parameters for in-distribution (ID) data to reduce the negative impact of redundant parameters on OOD detection. However, we empirically find that these selected parameters may behave overconfidently toward OOD data and hurt OOD detection. To address this issue, we propose a simple yet effective post-hoc method called Instance-aware Test Pruning (ITP), which performs OOD detection by considering both coarse-grained and fine-grained levels of parameter pruning. Specifically, ITP first estimates the class-specific parameter contribution distribution by exploring the ID data. By using the contribution distribution, ITP conducts coarse-grained pruning to eliminate redundant parameters. More importantly, ITP further adopts a fine-grained test pruning process based on the right-tailed Z-score test, which can adaptively remove instance-level overconfident parameters. Finally, ITP derives OOD scores from the pruned model to achieve more reliable predictions. Extensive experiments on widely adopted benchmarks verify the effectiveness of ITP, demonstrating its competitive performance.

Figures

Figures reproduced from arXiv: 2412.12566 by the authors.

**Figure 2.** Illustration of OOD detection using ITP. The overall procedure involves three main steps. (1) Training data are used to [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Changes in OOD score distribution using ITP on [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: The FFPR95 and AUROC with different number [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: The distribution of parameter contributions on ID data (CIFAR-10) using (a) DenseNet-101 and (b) ResNet-18 as [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 37 canonical work pages

[1]

L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al

Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al. 2023. Gpt-4 technical report. CoRR, abs/2303.08774

arXiv 2023
[2]

H.; Park, G.; and Kim, S

Ahn, Y. H.; Park, G.; and Kim, S. T. 2023. LINe: Out-of-Distribution Detection by Leveraging Important Neurons. In CVPR, pages 19852--19862

work page 2023
[3]

L.; Li, X.; He, W.; Gou, L.; and Ren, L

Behpour, S.; Doan, T. L.; Li, X.; He, W.; Gou, L.; and Ren, L. 2023. GradOrth: A Simple yet Efficient Out-of-Distribution Detection with Orthogonal Projection of Gradients. In NeurIPS

work page 2023
[4]

Bendale, A.; and Boult, T. E. 2016. Towards Open Set Deep Networks. In CVPR, pages 1563--1572

work page 2016
[5]

Chen, C.; Fu, Z.; Liu, K.; Chen, Z.; Tao, M.; and Ye, J. 2023. Optimal Parameter and Neuron Pruning for Out-of-Distribution Detection. In NeurIPS

work page 2023
[6]

Cimpoi, M.; Maji, S.; Kokkinos, I.; Mohamed, S.; and Vedaldi, A. 2014. Describing Textures in the Wild. In CVPR, pages 3606--3613

work page 2014
[7]

Djurisic, A.; Bozanic, N.; Ashok, A.; and Liu, R. 2023. Extremely Simple Activation Shaping for Out-of-Distribution Detection. In ICLR

work page 2023
[8]

Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; Uszkoreit, J.; and Houlsby, N. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In ICLR

2021

Show all 47 references

[9]

Geiger, A.; Lenz, P.; and Urtasun, R. 2012. Are we ready for autonomous driving? The KITTI vision benchmark suite. In CVPR, pages 3354--3361

2012
[10]

S.; Sun, Y.; and Li, Y

Ghosal, S. S.; Sun, Y.; and Li, Y. 2024. How to Overcome Curse-of-Dimensionality for Out-of-Distribution Detection? In AAAI, pages 19849--19857

2024
[11]

He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep Residual Learning for Image Recognition. In CVPR, pages 770--778

2016
[12]

He, R.; Yuan, Y.; Han, Z.; Wang, F.; Su, W.; Yin, Y.; Liu, T.; and Gong, Y. 2024. Exploring Channel-Aware Typical Features for Out-of-Distribution Detection. In AAAI, pages 12402--12410

2024
[13]

Hendrycks, D.; Basart, S.; Mazeika, M.; Zou, A.; Kwon, J.; Mostajabi, M.; Steinhardt, J.; and Song, D. 2022. Scaling Out-of-Distribution Detection for Real-World Settings. In ICML, volume 162, pages 8759--8773

2022
[14]

Hendrycks, D.; and Gimpel, K. 2017. A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks. In ICLR

2017
[15]

V.; Aodha, O

Horn, G. V.; Aodha, O. M.; Song, Y.; Cui, Y.; Sun, C.; Shepard, A.; Adam, H.; Perona, P.; and Belongie, S. J. 2018. The INaturalist Species Classification and Detection Dataset. In CVPR, pages 8769--8778

2018
[16]

Huang, G.; Liu, Z.; van der Maaten, L.; and Weinberger, K. Q. 2017. Densely Connected Convolutional Networks. In CVPR, pages 2261--2269

2017
[17]

Huang, H.; Li, Z.; Wang, L.; Chen, S.; Zhou, X.; and Dong, B. 2021. Feature Space Singularity for Out-of-Distribution Detection. In AAAI, volume 2808

2021
[18]

Huang, R.; Geng, A.; and Li, Y. 2021. On the Importance of Gradients for Detecting Distributional Shifts in the Wild. In NeurIPS, pages 677--689

2021
[19]

Kong, H.; and Li, H. 2022. BFAct: Out-of-Distribution Detection with Butterworth Filter Rectified Activations. In Sun, F.; Cangelosi, A.; Zhang, J.; Yu, Y.; Liu, H.; and Fang, B., eds., ICCSIP, volume 1787, pages 115--129

2022
[20]

Krizhevsky, A. 2009. Learning Multiple Layers of Features from Tiny Images

2009
[21]

Lee, K.; Lee, K.; Lee, H.; and Shin, J. 2018. A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks. In NeurIPS, pages 7167--7177

2018
[22]

Liang, S.; Li, Y.; and Srikant, R. 2018. Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks. In ICLR

2018
[23]

E.; Setio, A

Litjens, G.; Kooi, T.; Bejnordi, B. E.; Setio, A. A. A.; Ciompi, F.; Ghafoorian, M.; Van Der Laak, J. A.; Van Ginneken, B.; and S \'a nchez, C. I. 2017. A survey on deep learning in medical image analysis. Medical image analysis, 42: pages 60--88

2017
[24]

D.; and Li, Y

Liu, W.; Wang, X.; Owens, J. D.; and Li, Y. 2020. Energy-based Out-of-distribution Detection. In NeurIPS

2020
[25]

Malinin, A.; and Gales, M. J. F. 2018. Predictive Uncertainty Estimation via Prior Networks. In NeurIPS, pages 7047--7058

2018
[26]

H.; and V \' a zquez, D

Monteiro, J.; Rodr \' guez, P.; No \" e l, P.; Laradji, I. H.; and V \' a zquez, D. 2023. Constraining Representations Yields Models That Know What They Don't Know. In ICLR

2023
[27]

Netzer, Y.; Wang, T.; Coates, A.; Bissacco, A.; Wu, B.; and Ng, A. 2011. Reading Digits in Natural Images with Unsupervised Feature Learning

2011
[28]

Radford, A.; Narasimhan, K.; Salimans, T.; Sutskever, I.; et al. 2018. Improving language understanding by generative pre-training

2018
[29]

S.; et al

Shapley, L. S.; et al. 1953. A value for n-person games. Classics in game theory

1953
[30]

Sun, Y.; Guo, C.; and Li, Y. 2021. ReAct: Out-of-distribution Detection With Rectified Activations. In NeurIPS, pages 144--157

2021
[31]

Sun, Y.; and Li, Y. 2022. DICE: Leveraging Sparsification for Out-of-Distribution Detection. In ECCV, volume 13684, pages 691--708

2022
[32]

Sun, Y.; Ming, Y.; Zhu, X.; and Li, Y. 2022. Out-of-Distribution Detection with Deep Nearest Neighbors. In ICML, volume 162, pages 20827--20840

2022
[33]

Wang, H.; Li, Z.; Feng, L.; and Zhang, W. 2022. ViM: Out-Of-Distribution with Virtual-logit Matching. In CVPR, pages 4911--4920

2022
[34]

A.; Oliva, A.; and Torralba, A

Xiao, J.; Hays, J.; Ehinger, K. A.; Oliva, A.; and Torralba, A. 2010. SUN database: Large-scale scene recognition from abbey to zoo. In CVPR, pages 3485--3492

2010
[35]

Xu, M.; Lian, Z.; Liu, B.; and Tao, J. 2023. VRA: Variational Rectified Activation for Out-of-distribution Detection. In NeurIPS

2023
[36]

A.; Zhang, Y.; Finkelstein, A.; Kulkarni, S

Xu, P.; Ehinger, K. A.; Zhang, Y.; Finkelstein, A.; Kulkarni, S. R.; and Xiao, J. 2015. Turkergaze: Crowdsourcing saliency with webcam based eye tracking. CoRR, abs/1504.06755

2015 arXiv
[37]

Yang, J.; Wang, P.; Zou, D.; Zhou, Z.; Ding, K.; Peng, W.; Wang, H.; Chen, G.; Li, B.; Sun, Y.; Du, X.; Zhou, K.; Zhang, W.; Hendrycks, D.; Li, Y.; and Liu, Z. 2022. OpenOOD: Benchmarking Generalized Out-of-Distribution Detection. In NeurIPS

2022
[38]

Yang, Y.; Huang, Y.; Guo, W.; Xu, B.; and Xia, D. 2023 a . Towards Global Video Scene Segmentation with Context-Aware Transformer. In AAAI, 3206--3213

2023
[39]

Yang, Y.; Jiang, N.; Xu, Y.; and Zhan, D. 2024. Robust Semi-Supervised Learning by Wisely Leveraging Open-Set Data. TPAMI, 46(12): 8334--8347

2024
[40]

Yang, Y.; Zhang, C.; Xu, Y.; Yu, D.; Zhan, D.; and Yang, J. 2021. Rethinking Label-Wise Cross-Modal Retrieval from A Semantic Sharing Perspective. In IJCAI, 3300--3306

2021
[41]

Yang, Y.; Zhang, Y.; Song, X.; and Xu, Y. 2023 b . Not All Out-of-Distribution Data Are Harmful to Open-Set Active Learning. In Oh, A.; Naumann, T.; Globerson, A.; Saenko, K.; Hardt, M.; and Levine, S., eds., NeurIPS

2023
[42]

Yu, F.; Zhang, Y.; Song, S.; Seff, A.; and Xiao, J. 2015. LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop. CoRR, abs/1506.03365

2015 arXiv
[43]

Zhang, J.; Yang, J.; Wang, P.; Wang, H.; Lin, Y.; Zhang, H.; Sun, Y.; Du, X.; Zhou, K.; Zhang, W.; Li, Y.; Liu, Z.; Chen, Y.; and Li, H. 2023. OpenOOD v1.5: Enhanced Benchmark for Out-of-Distribution Detection. CoRR, abs/2306.09301

2023 arXiv
[44]

Zhou, B.; Lapedriza, \` A .; Khosla, A.; Oliva, A.; and Torralba, A. 2018. Places: A 10 Million Image Database for Scene Recognition. TPAMI

2018
[45]

Zhu, Y.; Chen, Y.; Xie, C.; Li, X.; Zhang, R.; Xue, H.; Tian, X.; Zheng, B.; and Chen, Y. 2022. Boosting Out-of-distribution Detection with Typical Features. In NeurIPS

2022
[46]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...
[47]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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

[1] [1]

L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al

Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al. 2023. Gpt-4 technical report. CoRR, abs/2303.08774

arXiv 2023

[2] [2]

H.; Park, G.; and Kim, S

Ahn, Y. H.; Park, G.; and Kim, S. T. 2023. LINe: Out-of-Distribution Detection by Leveraging Important Neurons. In CVPR, pages 19852--19862

work page 2023

[3] [3]

L.; Li, X.; He, W.; Gou, L.; and Ren, L

Behpour, S.; Doan, T. L.; Li, X.; He, W.; Gou, L.; and Ren, L. 2023. GradOrth: A Simple yet Efficient Out-of-Distribution Detection with Orthogonal Projection of Gradients. In NeurIPS

work page 2023

[4] [4]

Bendale, A.; and Boult, T. E. 2016. Towards Open Set Deep Networks. In CVPR, pages 1563--1572

work page 2016

[5] [5]

Chen, C.; Fu, Z.; Liu, K.; Chen, Z.; Tao, M.; and Ye, J. 2023. Optimal Parameter and Neuron Pruning for Out-of-Distribution Detection. In NeurIPS

work page 2023

[6] [6]

Cimpoi, M.; Maji, S.; Kokkinos, I.; Mohamed, S.; and Vedaldi, A. 2014. Describing Textures in the Wild. In CVPR, pages 3606--3613

work page 2014

[7] [7]

Djurisic, A.; Bozanic, N.; Ashok, A.; and Liu, R. 2023. Extremely Simple Activation Shaping for Out-of-Distribution Detection. In ICLR

work page 2023

[8] [8]

Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; Uszkoreit, J.; and Houlsby, N. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In ICLR

2021