GARD cuts CLIP visual backdoor success below 1%
The post-hoc defense covers six visual backdoor attacks on two CLIP backbones and needs no extra inference compute.
Stories per day · 30 days
2 RSS
Relatedmultimodal4clip2embeddings2geospatial2location-encoders2attention-sink1backdoor1biomedical1clap1cross-architecture1hard-negatives1information-retrieval1instruction-following1knowledge-distillation1knowledge-retrieval1landmark-detection1llm1medical-imaging1
24 stories, newest first · one card per story
Stories tagged #contrastive-learning
Trainable [DUMP] prefix tokens cut interference in retrieval embeddings
The method leaves architecture, pooling, and contrastive loss unchanged and reports gains on retrieval and long-context benchmarks.
Pixel contrastive learning improves spine landmark detection on X-rays
A U-Net matches dense pixel embeddings to landmark prototypes, then uses learnable shape priors to resolve ambiguous vertebral corners.
NHNM mines natural hard negatives for instruction retrieval
Fine-tuning 568M to 4B models on nearly 7K NHNM-MARCO instances matches or beats much larger retrievers on InstructIR and FollowIR.
LoCA aligns location encoders with multimodal satellite imagery
On 17 tasks LoCA has the most wins and is the only model ahead on a majority of tasks.
Simplex embeddings cut the multimodal modality gap by 97–99%
Softmax adapters and Total Variation similarity match or beat unconstrained retrieval on five benchmarks.
LoCA co-aligns Sentinel encoders for geospatial location encoding
On 17 tasks LoCA wins 68.8% of per-task comparisons and is the only encoder ahead of every rival on a majority of tasks.
CoRE learns rule routing and constrained retrieval without labeled pairs
A router and rule-constrained retriever co-adapt with reinforcement and contrastive learning on STaRK-QA.
SimOPD: Relational Knowledge Transfer for On-Policy Distillation
SimOPD distills teacher knowledge through response similarities, gaining up to 3.1 points cross-architecture.
Global Average Precision for Representation Learning
Researchers introduce gSAP, a global average precision loss that can replace InfoNCE in representation learning.
Learning Multimodal Embeddings with Evidence-Aligned Readout
EviAlign reads multimodal evidence at semantic boundaries, reaching 76.9 Recall@1 across 12 MMEB tasks.
Transfer Learning for Edge Classification on Dynamic Text-Attributed Graphs
Researchers show dynamic graph models transfer poorly and propose STSA for cross-domain edge classification.
SynCo: Learning Cross-Modal Synergy by Contrasting Interaction Residuals
SynCo trains multimodal synergy with contrastive loss on interaction residuals, gaining 5.98 points on Trifeature.
Contrastive-LM Releases CLM-8B: An Open System One Model That Scores Agent Actions Up to 9× Faster Than Jev
Contrastive-LM released CLM-8B, an open model that scores agent actions up to 9× faster than Jev.
4
Contrastive Language Models
A Hacker News post on Contrastive Language Models has 41 points and six comments, with no technical details provided.
Contrastive Learning for Authorship Verification
A ModernBERT bi-encoder using contrastive learning reaches 98.4% accuracy on PAN21 authorship verification.
Discovery-Driven Integration of Disjoint Tables via Text
LOKI learns row-sentence links to join disjoint data-lake tables, with 0.982 precision and far lower LLM cost.
HYDRA: Proactive Android Malware Drift Adaptation via Hierarchical Graph Contrastive Learning
HYDRA uses hierarchical Android app graphs and contrastive learning to adapt malware detectors to concept drift with far fewer labels.
Contrastive-LM/CLM-v0.1-8B — new model trending #30 on Hugging Face
Contrastive-LM released CLM-v0.1-8B, a Qwen3-8B contrastive model for fast state-action scoring and agent verification.
2
ToneCL: Contrastive Learning for Few-Shot Syllable-Level Tone Classification
ToneCL reaches 91.6% few-shot Mandarin tone accuracy using contrastive pretraining on unlabeled speech.
Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings
Ovis-Embedding reports state-of-the-art omni-modal retrieval across text, image, video, and audio.
FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation
FLAT jointly trains a multimodal encoder with text-to-image and image-to-text decoders, producing flexible-length tokens that hit 83.1 GenEval on T2I after fine-tuning.
Anatomical Grounding and Leakage-Aware Multimodal Contrastive Learning for Alzheimer's Disease Classification from Structural MRI
Study of label leakage and anatomical grounding in multimodal MRI models for Alzheimer's staging shows cognitive-score fusion accuracy of 87.3% is leakage-driven.
ReMoMask-2: Latent Retrieval-Augmented Masked Motion Generation
ReMoMask-2 rebuilds retrieval in the generator's latent space for text-to-motion generation, achieving lowest FID on KIT-ML and SnapMoGen.