Similar Items: How Many Iterations to Jailbreak? Dynamic Budget Allocation for Multi-Turn LLM Evaluation
- Continual Knowledge Updating in LLM Systems: Learning Through Multi-Timescale Memory Dynamics
- DARTS: Targeting Prognostic Covariates in Budget-Constrained Sequential Experiments
- Exact ReLU realization of tensor-product refinement iterates
- Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction
- Adaptive Policy Selection and Fine-Tuning under Interaction Budgets for Offline-to-Online Reinforcement Learning
- Evaluating the Architectural Reasoning Capabilities of LLM Provers via the Obfuscated Natural Number Game