Why Fixed Linear Steering Fails in Medical LLMs
Explore why fixed linear steering can't correct failures in medical LLMs and the impact of representational entanglement on AI model performance.
Exploiting Reconstruction-Concealment Tradeoff in MLLMs
Discover new methods to exploit the reconstruction-concealment tradeoff in multimodal LLMs, revealing vulnerabilities and challenges in AI safety protocols...
Stochastic Causal Learning for Precision Medicine Accuracy
Discover how stochastic causal representation learning resolves bias-precision paradox, improving treatment predictions and clinical decision support.
Enhancing Self-Evolving Search Agents with Knowledge-Graph Paths
Discover how knowledge-graph paths improve self-evolving search agents by boosting question quality and reward feedback in AI training.
Inference-Time Budget Control for Efficient LLM Search Agents
Optimize LLM search agents with inference-time budget control to enhance multi-hop QA accuracy and efficiency under strict resource limits.