performance drop
A decrease in the accuracy or effectiveness of an AI model, often observed when the model is applied to new conditions or datasets, raising concerns about generalizability and robustness.
- Adaptive Distraction: Probing LLM Contextual Robustness with Automated Tree Search
- Beyond Single-Task: Robust Multi-Task Length Generalization for LLMs
- Ineq-Comp: Benchmarking Human-Intuitive Compositional Reasoning in Automated Theorem Proving of Inequalities
- Mind the GAP! The Challenges of Scale in Pixel-based Deep Reinforcement Learning
- Mind the Gap: Removing the Discretization Gap in Differentiable Logic Gate Networks
- Training-Free Efficient Video Generation via Dynamic Token Carving