inference cost
Inference cost refers to the computational resources and time required to make predictions or execute a model after it has been trained. Reducing inference costs is important for deploying AI systems in production, where efficiency directly impacts user experience and scalability.
- AcuRank: Uncertainty-Aware Adaptive Computation for Listwise Reranking
- AgentTTS: Large Language Model Agent for Test-time Compute-optimal Scaling Strategy in Complex Tasks
- Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency
- FreeControl: Efficient, Training-Free Structural Control via One-Step Attention Extraction
- Leaving No OOD Instance Behind: Instance-Level OOD Fine-Tuning for Anomaly Segmentation
- NFIG: Multi-Scale Autoregressive Image Generation via Frequency Ordering
- ODG: Occupancy Prediction Using Dual Gaussians
- PrefixKV: Adaptive Prefix KV Cache is What Vision Instruction-Following Models Need for Efficient Generation
- SCAN: Self-Denoising Monte Carlo Annotation for Robust Process Reward Learning
- Towards Better & Faster Autoregressive Image Generation: From the Perspective of Entropy