AgentPulse: Continuous AI Agent Evaluation Framework
AgentPulse offers a dynamic framework to continuously evaluate AI agents using multi-signal data for real-world performance and adoption insights.
QED: Open-Source AI System for Mathematical Proofs
Discover QED, an open-source multi-agent AI system that generates original mathematical proofs for open problems with expert validation.
CT-FineBench: Benchmark for Accurate CT Report Evaluation
CT-FineBench offers a fine-grained benchmark to evaluate CT report generation, improving diagnostic accuracy and aligning with expert clinical assessments.
Failure-Focused Evaluation for Trilingual Public AI Agents
Discover PSA-Eval, a failure-centered framework improving runtime evaluation of deployed trilingual public-space AI agents for enhanced reliability.
How Representational Curvature Affects Uncertainty in LLMs
Explore how representational curvature influences behavioral uncertainty and token prediction accuracy in large language models like GPT-2 and Pythia.