Pith. sign in

REVIEW 10 cited by

Contrastive Learning with Hard Negative Samples

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

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 2010.04592 v2 pith:SVDFBL5Y submitted 2020-10-09 cs.LG stat.ML

Joshua Robinson , Ching-Yao Chuang , Suvrit Sra , Stefanie Jegelka This is my paper
classification cs.LGstat.ML
keywords negativecontrastivehardlearningsamplessamplingmethodsunsupervised
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

How can you sample good negative examples for contrastive learning? We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point). The key challenge toward using hard negatives is that contrastive methods must remain unsupervised, making it infeasible to adopt existing negative sampling strategies that use true similarity information. In response, we develop a new family of unsupervised sampling methods for selecting hard negative samples where the user can control the hardness. A limiting case of this sampling results in a representation that tightly clusters each class, and pushes different classes as far apart as possible. The proposed method improves downstream performance across multiple modalities, requires only few additional lines of code to implement, and introduces no computational overhead.

Discussion (0). Sign in to comment.

Forward citations

Cited by 10 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 37 citations worldwide. Full citation record

  1. Fusion Embedding: A Unified Embedding Space for Text, Image, Video, and Audio

    cs.CL 2026-07 conditional novelty 7.0 of 10

    Trained connectors and audio-only gated adapters integrate audio into a frozen vision-language embedding space, preserving base outputs bit-exactly and yielding emergent audio-image retrieval.

  2. DisProtEdit: Exploring Disentangled Representations for Multi-Attribute Protein Editing

    q-bio.QM 2025-06 conditional novelty 7.0 of 10

    DisProtEdit learns disentangled protein representations from separate structural and functional text descriptions, enabling controllable single- and multi-attribute protein editing via latent interpolation.

  3. Discriminative Axis, Not Data Volume: What a Contrastive Corpus Teaches an Audio Embedding

    cs.CL 2026-08 conditional novelty 6.0 of 10

    A contrastive audio embedding learns an attribute only when in-batch negatives cannot be separated without it, so corpus structure, not size or caption vocabulary, controls what is encoded.

  4. Generalizable Object Re-Identification via Visual In-Context Prompting

    cs.CV 2025-08 conditional novelty 6.0 of 10

    VICP uses an LLM to generate per-category visual prompts for a frozen DINOv2, enabling few-shot generalization to unseen object categories in re-identification without parameter updates.

  5. QuRe: Query-Relevant Retrieval through Hard Negative Sampling in Composed Image Retrieval

    cs.CV 2025-07 conditional novelty 6.0 of 10

    QuRe trains composed image retrieval models with a pairwise reward objective on hard negatives found between sharp relevance-score drops, and adds a human-preference benchmark for evaluating retrieval relevance.

  6. Multimodal Semantic-Aware Contrastive Learning For False Negative Mitigation in 3D Medical Imaging

    cs.LG 2026-07 conditional novelty 5.0 of 10

    MseaCL, a semantic-aware contrastive pretraining method for 3D brain MRI and reports, reports a 0.226 external AUC gain over instance-based contrastive learning for pediatric brain tumor molecular classification.

  7. A Generalized Learning Framework for Self-Supervised Contrastive Learning

    cs.LG 2025-08 unverdicted novelty 5.0 of 10

    A single framework unifies BYOL, Barlow Twins, and SwAV, plus a plug-in calibration method, ADC, that improves learned representations by preserving input-space distances.

  8. The Impact of Semantic Pairs on Self-Supervised Representation Learning

    cs.LG 2025-10 conditional novelty 4.0 of 10

    Semantic positive pairs (same-class, different images) consistently beat augmented-view pairs for self-supervised pretraining in controlled ImageNet-subset experiments, with the largest gains for SimCLR.

  9. eMargin: Revisiting Contrastive Learning with Margin-Based Separation

    cs.LG 2025-07 reject novelty 4.0 of 10

    An adaptive margin added to InfoNCE improves time series clustering metrics but hurts linear-probe classification, exposing a disconnect between clustering scores and downstream utility.

  10. Subject Invariant Contrastive Learning for Human Activity Recognition

    cs.CV 2025-07 conditional novelty 4.0 of 10

    A subject-reweighted contrastive loss improves cross-subject generalization for human activity recognition across unimodal, multimodal, and supervised settings.

Pith tools