model-agnostic
Refers to methods or algorithms that can be applied across various model architectures without being specific to any one type, allowing for flexibility and broader applicability in the analysis or optimization of machine learning models.
- Accelerating Feature Conformal Prediction via Taylor Approximation
- Are Pixel-Wise Metrics Reliable for Computerized Tomography Reconstruction?
- Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs
- Conformal Prediction for Causal Effects of Continuous Treatments
- DenoiseRotator: Enhance Pruning Robustness for LLMs via Importance Concentration
- Document Summarization with Conformal Importance Guarantees
- Dynamic and Chemical Constraints to Enhance the Molecular Masked Graph Autoencoders
- Dynamical Low-Rank Compression of Neural Networks with Robustness under Adversarial Attacks
- Dynamical Low-Rank Compression of Neural Networks with Robustness under Adversarial Attacks
- Enhancing Graph Classification Robustness with Singular Pooling
- FACE: A General Framework for Mapping Collaborative Filtering Embeddings into LLM Tokens
- Feature Distillation is the Better Choice for Model-Heterogeneous Federated Learning
- Glocal Information Bottleneck for Time Series Imputation
- LayerIF: Estimating Layer Quality for Large Language Models using Influence Functions
- Negative Feedback Really Matters: Signed Dual-Channel Graph Contrastive Learning Framework for Recommendation
- Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models
- Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data
- Post Hoc Regression Refinement via Pairwise Rankings
- Probabilistic Stability Guarantees for Feature Attributions
- SRA-CL: Semantic Retrieval Augmented Contrastive Learning for Sequential Recommendation
- STAR: Spatial-Temporal Tracklet Matching for Multi-Object Tracking
- SceneForge: Enhancing 3D-text alignment with Structured Scene Compositions
- Uncertainty-Informed Meta Pseudo Labeling for Surrogate Modeling with Limited Labeled Data