inference time
Inference time is the duration taken by an AI system to make predictions once the model has been trained. It is a vital consideration in deployment, as quicker inference allows for real-time applications, while longer times may limit usability.
- Activation Control for Efficiently Eliciting Long Chain-of-thought Ability of Language Models
- AgentDAM: Privacy Leakage Evaluation for Autonomous Web Agents
- AutoJudge: Judge Decoding Without Manual Annotation
- Broken Tokens? Your Language Model can Secretly Handle Non-Canonical Tokenizations
- Cost-Aware Contrastive Routing for LLMs
- Depth-Width Tradeoffs for Transformers on Graph Tasks
- Efficient and Generalizable Mixed-Precision Quantization via Topological Entropy
- GLVD: Guided Learned Vertex Descent
- Incentivizing LLMs to Self-Verify Their Answers
- Language Ranker: A Lightweight Ranking framework for LLM Decoding
- Learning conformational ensembles of proteins based on backbone geometry
- LinEAS: End-to-end Learning of Activation Steering with a Distributional Loss
- Memory Mosaics at scale
- Memory Mosaics at scale
- Repurposing Marigold for Zero-Shot Metric Depth Estimation via Defocus Blur Cues
- VL-SAM-V2: Open-World Object Detection with General and Specific Query Fusion