downstream performance
Downstream performance refers to the effectiveness of an AI model when applied to specific tasks or applications after training. It reflects how well a model generalizes to new, potentially unseen data in the real world and is a critical measure of its utility.
- 3D-GSRD: 3D Molecular Graph Auto-Encoder with Selective Re-mask Decoding
- Block-Diagonal LoRA for Eliminating Communication Overhead in Tensor Parallel LoRA Serving
- Brain-tuning Improves Generalizability and Efficiency of Brain Alignment in Speech Models
- Diffusion Beats Autoregressive in Data-Constrained Settings
- Does Object Binding Naturally Emerge in Large Pretrained Vision Transformers?
- Exploring Structural Degradation in Dense Representations for Self-supervised Learning
- From Faults to Features: Pretraining to Learn Robust Representations against Sensor Failures
- Panacea: Mitigating Harmful Fine-tuning for Large Language Models via Post-fine-tuning Perturbation
- PiKE: Adaptive Data Mixing for Large-Scale Multi-Task Learning Under Low Gradient Conflicts
- Randomized-MLP Regularization Improves Domain Adaptation and Interpretability in DINOv2
- Reverse Engineering Human Preferences with Reinforcement Learning
- SAM2Flow: Interactive Optical Flow Estimation with Dual Memory for in vivo Microcirculation Analysis
- Self-Adapting Language Models
- THUNDER: Tile-level Histopathology image UNDERstanding benchmark
- The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness
- URLs Help, Topics Guide: Understanding Metadata Utility in LLM Training
- Win Fast or Lose Slow: Balancing Speed and Accuracy in Latency-Sensitive Decisions of LLMs