Pith. sign in

REVIEW 4 cited by

Active Learning by Acquiring Contrastive Examples

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2109.03764 v1 pith:WXM3UFDF submitted 2021-09-08 cs.CL cs.AIcs.LG

Active Learning by Acquiring Contrastive Examples

classification cs.CL cs.AIcs.LG
keywords dataacquisitionactivecontrastivelearningbestbetterdatasets
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Common acquisition functions for active learning use either uncertainty or diversity sampling, aiming to select difficult and diverse data points from the pool of unlabeled data, respectively. In this work, leveraging the best of both worlds, we propose an acquisition function that opts for selecting \textit{contrastive examples}, i.e. data points that are similar in the model feature space and yet the model outputs maximally different predictive likelihoods. We compare our approach, CAL (Contrastive Active Learning), with a diverse set of acquisition functions in four natural language understanding tasks and seven datasets. Our experiments show that CAL performs consistently better or equal than the best performing baseline across all tasks, on both in-domain and out-of-domain data. We also conduct an extensive ablation study of our method and we further analyze all actively acquired datasets showing that CAL achieves a better trade-off between uncertainty and diversity compared to other strategies.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Multimodal Distribution Matching for Vision-Language Dataset Distillation

    cs.CV 2026-05 unverdicted novelty 6.0

    MDM distills vision-language datasets via joint embedding clustering, weight-space model interpolation, and geometry-aware distribution matching on the unit hypersphere.

  2. OrderDP: A Theoretically Guaranteed Lossless Dynamic Data Pruning Framework

    cs.LG 2026-06 unverdicted novelty 5.0

    OrderDP is a plug-and-play data pruning method that selects a random subset then top-q samples to guarantee unbiased surrogate-loss training with convergence analysis and over 40% training cost reduction on CIFAR and ...

  3. Labeled TrustSet Guided: Batch Active Learning with Reinforcement Learning

    cs.LG 2026-04 unverdicted novelty 5.0

    BRAL-T uses TrustSet-guided reinforcement learning for batch active learning and reports state-of-the-art results on 10 image classification benchmarks plus 2 fine-tuning tasks.

  4. \emph{FoQuS}: A Forgetting-Quality Coreset Selection Framework for Automatic Modulation Recognition

    cs.LG 2025-09 reject novelty 5.0

    FoQuS uses three training-dynamics scores to select a 1% to 30% coreset for AMR and claims higher accuracy than ten existing coreset methods.