parameter efficiency
The ability of a model to achieve high performance with a minimal number of parameters, offering advantages in terms of speed and resource usage.
- AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play
- Block-Diagonal LoRA for Eliminating Communication Overhead in Tensor Parallel LoRA Serving
- Boosting Resilience of Large Language Models through Causality-Driven Robust Optimization
- CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching
- Composing Linear Layers from Irreducibles
- Corrector Sampling in Language Models
- FedMGP: Personalized Federated Learning with Multi-Group Text-Visual Prompts
- Fourier Analysis Network
- GaRA-SAM: Robustifying Segment Anything Model with Gated-Rank Adaptation
- Learning to Factorize Spatio-Temporal Foundation Models
- LoRA vs Full Fine-tuning: An Illusion of Equivalence
- MOSDT: Self-Distillation-Based Decision Transformer for Multi-Agent Offline Safe Reinforcement Learning
- MixPrompt: Efficient Mixed Prompting for Multimodal Semantic Segmentation
- Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation
- Nemotron-Flash: Towards Latency-Optimal Hybrid Small Language Models
- On the Integration of Spatial-Temporal Knowledge: A Lightweight Approach to Atmospheric Time Series Forecasting
- Optimal Control for Transformer Architectures: Enhancing Generalization, Robustness and Efficiency
- Pan-LUT: Efficient Pan-sharpening via Learnable Look-Up Tables
- Pan-LUT: Efficient Pan-sharpening via Learnable Look-Up Tables
- PreFM: Online Audio-Visual Event Parsing via Predictive Future Modeling
- Revolutionizing Graph Aggregation: From Suppression to Amplification via BoostGCN
- S$^2$M-Former: Spiking Symmetric Mixing Branchformer for Brain Auditory Attention Detection
- S'MoRE: Structural Mixture of Residual Experts for Parameter-Efficient LLM Fine-tuning
- Segment Anything Model Meets Semi-supervised Medical Image Segmentation: A Novel Perspective
- Tensor-Parallelism with Partially Synchronized Activations
- The Rise of Parameter Specialization for Knowledge Storage in Large Language Models
- Unified Scaling Laws for Compressed Representations