Representation Learning

Standard Overall results available

MantisV2

A more lightweight Mantis encoder trained for stronger zero-shot time series representations.

Using MantisV2 for time series classification

A more lightweight Mantis encoder trained for stronger zero-shot time series representations.

In this benchmark, MantisV2 is evaluated using frozen representation + fitted classifier. Checkpoint information: MantisV2. Check the results below for the settings and dataset scopes with published scores.

Evaluated configurations

ScaledLR

Classifier heads are grouped under their shared pretrained model here. They remain separate leaderboard entries because they are different experimental configurations.

Organization: Huawei Noah’s Ark Lab · Université Paris Cité
Parameters: 4.19M backbone
Checkpoint used: MantisV2
Input Type: Univariate + multivariate time series
Inference: Frozen representation + fitted classifier

MantisV2 classification benchmark results

Compare accuracy and average rank only within the same evaluation setting and dataset scope. Standard uses the official training splits; 5-shot and 10-shot use different budget-feasible subsets. Rank and average rank include other models in the full comparison set. Scores are not averaged across configurations.

For comparisons with other backbones, open the foundation model classification leaderboard and use the same setting and scope. The few-shot evaluation protocol explains the label budgets.

Standard · Overall

Comparison set: 12 configurations

This Standard time series classification comparison reports 2 MantisV2 configurations for the overall dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 77.73%, reported by MantisV2+LR with coverage 198 / 198. This configuration ranks 2 of 12 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — Standard Overall time series classification benchmark results
Configuration	Overall Rank	Overall Accuracy (%)	Avg. Rank	Coverage
MantisV2+LR	2	77.73	4.64	198 / 198
MantisV2+RF	4	76.62	5.42	198 / 198

Standard · Univariate

Comparison set: 19 configurations

This Standard time series classification comparison reports 2 MantisV2 configurations for the univariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 80.03%, reported by MantisV2+LR with coverage 128 / 128. This configuration ranks 2 of 19 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — Standard Univariate time series classification benchmark results
Configuration	Univariate Rank	Univariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+LR	2	80.03	7.09	128 / 128
MantisV2+RF	8	78.79	8.70	128 / 128

Standard · Multivariate

Comparison set: 12 configurations

This Standard time series classification comparison reports 2 MantisV2 configurations for the multivariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 73.51%, reported by MantisV2+LR with coverage 70 / 70. This configuration ranks 2 of 12 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — Standard Multivariate time series classification benchmark results
Configuration	Multivariate Rank	Multivariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+LR	2	73.51	4.61	70 / 70
MantisV2+RF	5	72.66	4.94	70 / 70

5-shot · Overall

Comparison set: 18 configurations · Budget-feasible Low-shot subset

This 5-shot time series classification comparison reports 2 MantisV2 configurations for the overall dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 66.33%, reported by MantisV2+ScaledLR with coverage 140 / 140. This configuration ranks 1 of 18 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 5-shot Overall time series classification benchmark results
Configuration	Overall Rank	Overall Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	1	66.33	7.21	140 / 140
MantisV2+RF	6	64.90	8.78	140 / 140

5-shot · Univariate

Comparison set: 19 configurations · Budget-feasible Low-shot subset

This 5-shot time series classification comparison reports 2 MantisV2 configurations for the univariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 67.54%, reported by MantisV2+ScaledLR with coverage 112 / 112. This configuration ranks 2 of 19 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 5-shot Univariate time series classification benchmark results
Configuration	Univariate Rank	Univariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	2	67.54	7.84	112 / 112
MantisV2+RF	7	66.09	9.57	112 / 112

5-shot · Multivariate

Comparison set: 18 configurations · Budget-feasible Low-shot subset

This 5-shot time series classification comparison reports 2 MantisV2 configurations for the multivariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 61.50%, reported by MantisV2+ScaledLR with coverage 28 / 28. This configuration ranks 6 of 18 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 5-shot Multivariate time series classification benchmark results
Configuration	Multivariate Rank	Multivariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	6	61.50	6.91	28 / 28
MantisV2+RF	7	60.15	8.38	28 / 28

10-shot · Overall

Comparison set: 18 configurations · Budget-feasible Low-shot subset

This 10-shot time series classification comparison reports 2 MantisV2 configurations for the overall dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 71.81%, reported by MantisV2+ScaledLR with coverage 121 / 121. This configuration ranks 3 of 18 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 10-shot Overall time series classification benchmark results
Configuration	Overall Rank	Overall Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	3	71.81	6.76	121 / 121
MantisV2+RF	6	70.03	8.60	121 / 121

10-shot · Univariate

Comparison set: 19 configurations · Budget-feasible Low-shot subset

This 10-shot time series classification comparison reports 2 MantisV2 configurations for the univariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 73.30%, reported by MantisV2+ScaledLR with coverage 97 / 97. This configuration ranks 3 of 19 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 10-shot Univariate time series classification benchmark results
Configuration	Univariate Rank	Univariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	3	73.30	7.17	97 / 97
MantisV2+RF	7	71.65	9.00	97 / 97

10-shot · Multivariate

Comparison set: 18 configurations · Budget-feasible Low-shot subset

This 10-shot time series classification comparison reports 2 MantisV2 configurations for the multivariate dataset scope. Each row shows its own coverage.

The highest reported average accuracy among this model's configurations in this setting is 65.80%, reported by MantisV2+ScaledLR with coverage 24 / 24. This configuration ranks 3 of 18 in this comparison set. The result describes this configuration and setting, rather than every possible use of the pretrained model.

MantisV2 — 10-shot Multivariate time series classification benchmark results
Configuration	Multivariate Rank	Multivariate Accuracy (%)	Avg. Rank	Coverage
MantisV2+ScaledLR	3	65.80	7.25	24 / 24
MantisV2+RF	11	63.49	9.67	24 / 24

MantisV2 paper, code and checkpoint

Reading this model's evaluation

The method catalog describes the pretrained model; leaderboard eligibility depends on the actual coverage of each experimental configuration. Catalog inclusion does not imply complete coverage of all 198 datasets.

Compare evaluated configurations Read the evaluation and ranking rules Explore aggregate capability analysis