Papers › MT-SLVR: Multi-Task Self-Supervised Learning for Transformation In(Variant) Representations

MT-SLVR: Multi-Task Self-Supervised Learning for Transformation In(Variant) Representations

29 May 2023arXiv:2305.17191archive 2025-07-28

Calum Heggan, Tim Hospedales, Sam Budgett, Mehrdad Yaghoobi

Contrastive self-supervised learning has gained attention for its ability to create high-quality representations from large unlabelled data sets. A key reason that these powerful features enable data-efficient learning of downstream tasks is that they provide augmentation invariance, which is often a useful inductive bias. However, the amount and type of invariances preferred is not known apriori, and varies across different downstream tasks. We therefore propose a multi-task self-supervised framework (MT-SLVR) that learns both variant and invariant features in a parameter-efficient manner. Our multi-task representation provides a strong and flexible feature that benefits diverse downstream tasks. We evaluate our approach on few-shot classification tasks drawn from a variety of audio domains and demonstrate improved classification performance on all of them

PaperPDFCode

Syntology This paper's arXiv id is not in Syntology's Atlas id list (549,449 ids, as this build read them; the read time is in the build record), so this page links to no Syntology page for it. The list is not the whole graph; Syntology's feed shows papers as they enter the graph.

Code

cheggan/mt-slvr officialmentioned in papermentioned on GitHubpytorchSyntology: not harvested report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Few-Shot Audio ClassificationInductive BiasSelf-Supervised LearningUnsupervised Few-Shot Audio Classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Few-Shot Audio Classification BirdClef 2020 (Pruned) SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 30.93±0.38 #8 of 10 Archive leaderboard report
Few-Shot Audio Classification BirdClef 2020 (Pruned) MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 29.49±0.38 #9 of 10 Archive leaderboard report
Few-Shot Audio Classification BirdClef 2020 (Pruned) Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 21.04±0.35 #10 of 10 Archive leaderboard report
Few-Shot Audio Classification CREMA-D MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 29.61±0.38 #1 of 3 Archive leaderboard report
Few-Shot Audio Classification CREMA-D SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 29.10±0.36 #2 of 3 Archive leaderboard report
Few-Shot Audio Classification CREMA-D Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 21.68±0.33 #3 of 3 Archive leaderboard report
Few-Shot Audio Classification Common Voice MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 35.22±0.40 #1 of 3 Archive leaderboard report
Few-Shot Audio Classification Common Voice SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 33.33±0.38 #2 of 3 Archive leaderboard report
Few-Shot Audio Classification Common Voice Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 23.00±0.42 #3 of 3 Archive leaderboard report
Few-Shot Audio Classification ESC-50 MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 69.53±0.39 #4 of 10 Archive leaderboard report
Few-Shot Audio Classification ESC-50 SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 63.40±0.39 #8 of 10 Archive leaderboard report
Few-Shot Audio Classification ESC-50 Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 37.76±0.34 #10 of 10 Archive leaderboard report
Few-Shot Audio Classification FSDKaggle2018 MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 39.11±0.41 #6 of 10 Archive leaderboard report
Few-Shot Audio Classification FSDKaggle2018 SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 37.64±0.40 #8 of 10 Archive leaderboard report
Few-Shot Audio Classification FSDKaggle2018 Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 21.72±0.34 #10 of 10 Archive leaderboard report
Few-Shot Audio Classification NSynth MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 71.81±0.39 #6 of 10 Archive leaderboard report
Few-Shot Audio Classification NSynth SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 66.44±0.40 #8 of 10 Archive leaderboard report
Few-Shot Audio Classification NSynth Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 62.52±0.36 #10 of 10 Archive leaderboard report
Few-Shot Audio Classification Speech Accent Archive MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 28.92±0.37 #1 of 3 Archive leaderboard report
Few-Shot Audio Classification Speech Accent Archive SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 26.16±0.34 #2 of 3 Archive leaderboard report
Few-Shot Audio Classification Speech Accent Archive Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 23.08±0.34 #3 of 3 Archive leaderboard report
Few-Shot Audio Classification Speech Command v2 SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 25.68±0.35 #1 of 3 Archive leaderboard report
Few-Shot Audio Classification Speech Command v2 MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 23.65±0.34 #2 of 3 Archive leaderboard report
Few-Shot Audio Classification Speech Command v2 Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 20.08±0.37 #3 of 3 Archive leaderboard report
Few-Shot Audio Classification VoxCeleb1 MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 33.58±0.39 #6 of 10 Archive leaderboard report
Few-Shot Audio Classification VoxCeleb1 SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 31.18±0.37 #7 of 10 Archive leaderboard report
Few-Shot Audio Classification VoxCeleb1 Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 21.68±0.40 #10 of 10 Archive leaderboard report
Few-Shot Audio Classification Watkins Marine Mammal Sounds MT-SLVR (SimCLR + MLAP) w/ Parallel Adapters (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 59.49±0.42 #1 of 5 Archive leaderboard report
Few-Shot Audio Classification Watkins Marine Mammal Sounds SimCLR (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 52.91±0.41 #3 of 5 Archive leaderboard report
Few-Shot Audio Classification Watkins Marine Mammal Sounds Multi-Label Augmentation Prediction (FSD50K, RN18) Top-1 Accuracy(5-Way-1-Shot) 28.88±0.39 #5 of 5 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

1x1 ConvolutionAverage PoolingBatch NormalizationBottleneck Residual BlockColorJitterConvolutionDense ConnectionsFeedforward NetworkGlobal Average PoolingKaiming InitializationMax PoolingNT-XentRandom Gaussian BlurRandom Resized CropReLUResidual BlockResidual ConnectionSimCLR

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections