#contrastive learning
32 resultsContrastive Reinforced Policy Optimization via Privileged Self-Distillation
Xingjian Wu, Junlin Liu, Xingchen Liu +6
The paper introduces Contrastive Reinforced Policy Optimization (CRPO), a method that frames on‑policy self‑distillation for large language models as a contrastive learning problem…
CDAE: Enhancing Perturbation Robustness in Pretrained Language Models with Contrastive Denoising
Sina Heydari, Amirreza Abbasi, Mohsen Hooshmand +1
The paper introduces a lightweight Contrastive Denoising Autoencoder (CDAE) that refines BERT sentence embeddings to be more robust against semantic-preserving perturbations by joi…
SKILL-KD: Contrastive Skill Distillation for LLM Agents
Qiming Shi, Yibo Dou, Jiawen Zhu +5
The paper introduces SKILL-KD, a contrastive skill distillation framework that creates explicit textual skill patches from teacher‑student failures to iteratively improve weaker LL…
SCALPEL: Semantic Cross-modal Alignment via LLM-Powered Encoder Learning for Medical Vision-Language Representation
Yunzhan Fu, Enyu Bao, Xiangyu Shen +4
The paper introduces SCALPEL, a framework that fine‑tunes a generative medical LLM into an isotropic encoder and uses an asymmetric, anatomy‑negation‑aware contrastive objective to…
Defending Against Backdoor Attacks via Alignment Checking in Model-Contrastive Federated Learning
Hongliang Zhang, Zhongyuan Yu, Guijuan Wang +4
The paper proposes FedDAB, a two‑phase defense for federated learning that uses contrastive regularization and alignment checking to detect and exclude malicious local updates caus…
KAMR: Grounding Generation via Knowledge-Aligned Multi-hop Retrieval
Xiaochen Wang, Yuan Zhong, Haoyu Wang +2
The paper presents KAMR, a knowledge‑aligned multi‑hop retriever that first identifies anchor graph triplets strongly tied to a query and then locally expands to connected evidence…
Zero-Fi: Zero-Shot Wi-Fi-Based Human Activity Recognition via Contrastive Signal-Language Alignment
Yitong Shen, Cheng Guo, Peiliang Wang +5
Zero-Fi introduces a contrastive learning framework that aligns Wi‑Fi signal features with natural‑language descriptions of activities, enabling recognition of unseen human activit…
R-SLPR: Region-based Small-to-Large Point-cloud Registration with Contrastive Learning
Yusen Wan, Zeyuan Chen, Qianshi Zou +1
The paper introduces R‑SLPR, a three‑stage framework that registers a small, partial point cloud to a much larger reference by proposing regions, matching them with contrastive lea…
Good Rankers, Bad Objectives: Bilinear Contrastive Critics under Expressive Policy Search
Ayushman Singh, Siddharth Aphale
The paper investigates why bilinear contrastive critics, which rank actions for reinforcement learning policies, can produce unsafe or misleading rankings due to issues like norm d…
Data Fusion and Contrastive Alignment for Unconstrained IR Molecular Structure Elucidation
Ethan J. Mick, Campbell A. Sweet, Matthias J. Young +1
The paper introduces a modified encoder‑decoder transformer with a Mixture‑of‑Experts decoder and contrastive alignment loss to predict full molecular structures directly from infr…
Multimodal Semantic-Aware Contrastive Learning For False Negative Mitigation in 3D Medical Imaging
Sara Ketabi, Matthias W. Wagner, Cynthia Hawkins +3
The paper proposes a multimodal semantic-aware contrastive learning framework that uses semantic similarity from radiology reports to reduce false negatives when training on 3D bra…
Angular Gaussian Supervised Contrastive Learning for Long-Tailed Electrocardiogram Arrhythmia Diagnosis
Jin Dai, Qiuzhen Zhang, Chenyun Dai +2
The paper introduces Angular Gaussian Supervised Contrastive Learning (AG‑SCL), a method that combines anisotropic contrastive embeddings, adaptive logit adjustment, and tail‑aware…
Similarity as Reward Alignment: Robust and Versatile Preference-based Reinforcement Learning
Sara Rajaram, R. James Cotton, Fabian H. Sinz
The paper proposes SARA, a contrastive method that learns latent representations of preferred behaviors and uses similarity as a reward signal, improving robustness to noisy human…
AspectCLIP: Optimizing CLIP Representation Space via Aspect-Guided Consistency Regularization
Yiyang Yao, Shanglin Liu, Jianming Lv +4
The paper introduces AspectCLIP, a method that groups image-text pairs by shared textual aspects and applies consistency regularization within these groups to avoid forcing unrelat…
Latent Trajectory Discrimination for AI-Generated Text Detection
Gianluca Bonifazi, Christopher Buratti, Michele Marchetti +5
The paper proposes a method that detects AI‑generated text by modeling how text representations change over the generation process, using trajectory‑based embeddings and contrastiv…
GlobalForge: Towards Robust AI-Generated Image Detection
Manni Cui, Ruiqi Liu, Dianyuan Zou +8
The paper introduces GlobalForge, a detection framework that shifts focus from fragile local artifacts to robust global structures to improve AI‑generated image detection under rea…
Towards Out-of-Distribution Detection in Vocoder Recognition via Latent Feature Reconstruction
Renmingyue Du, Jixun Yao, Qiuqiang Kong +1
The paper proposes a reconstruction‑based method using autoencoders to detect out‑of‑distribution vocoder samples by reconstructing WavLM acoustic features, with contrastive learni…
Contrastive Conformal Sets
Yahya Alkhatib, Wee Peng Tay
The paper introduces a method that combines contrastive learning with conformal prediction to create learnable geometric sets that guarantee a user‑specified coverage of positive s…
Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimization
Weiwen Xu, Jia Liu, Hou Pong Chan +4
The paper proposes Contrastive Policy Optimization, which leverages token‑level contrastive disagreement between reference‑guided and standard generation distributions to provide a…
Quality-Aware Robust Multi-View Clustering for Heterogeneous Observation Noise
Peihan Wu, Guanjie Cheng, Yufei Tong +2
The paper introduces QARMVC, a quality‑aware robust multi‑view clustering framework that estimates fine‑grained noise levels via reconstruction errors and uses instance‑level quali…
Personalizing Incremental Video Search with Hybrid Text and ID Embeddings
Vivek Kanojiya, Vishalaksh Aggarwal, Daeho Baek +2
The paper introduces a system for personalizing incremental video search on Apple TV by combining text‑based multilingual embeddings and ID‑based collaborative embeddings, using re…
The Hyperspherical Geometry of CLIP Latent Space: A Semantic Mixture Model
Zijie Yu, Gaowen Liu, Ramana Rao Kompella +2
The paper introduces a probabilistic model for CLIP embeddings using mixtures of von Mises-Fisher distributions on the unit hypersphere, improving density estimation and detection…
LaME: Learning to Think in Latent Space for Multimodal Embedding via Information Bottleneck
Peixi Wu, Biao Yang, Feipeng Ma +7
The paper introduces LaME, a multimodal embedding model that performs reasoning in a compact latent space using learnable tokens and an information‑bottleneck objective, eliminatin…
Contrastive-Augmented Flow Matching for Style-Content Disentanglement
Yusong Li, Pingchuan Ma, Ming Gui +2
The paper proposes Contrastive Augmented Flow Matching (CAtFM), a method that adds contrastive regularization to invertible flow matching to learn disentangled content and style re…