REAM: Optimizing Expert Pruning in Large Language Models
Discover how REAM improves pruning in Mixture-of-Experts LLMs by merging experts to boost performance and reduce memory use.
RoboPhD: Efficient Evolution of Complex AI Agents
Discover RoboPhD, a toolkit for evolving diverse AI agents efficiently under tight evaluation budgets, outperforming leading optimization methods.
Domain-Contextualized Inference: Efficient Graph Architecture
Discover a domain-focused graph architecture that optimizes inference with substrate-agnostic design and domain-scoped pruning for faster AI reasoning.
Safer Bargaining in LLM Agents with Surrogate Goals
Learn how surrogate goals improve safety in LLM-based agent bargaining by reducing risks through fine-tuning and scaffolding methods.
Thermodynamic GeoAI Reveals Regime Shifts in Spatial Systems
Discover how thermodynamic-inspired GeoAI uncovers regime-dependent mechanisms in heterogeneous spatial systems for better environmental and social insight...