JudgeMyAIAgentic Leaderboard Contamination: A Measurement-First Audit Framework
A deep dive into the measurement-first audit framework for agentic leaderboards, addressing contamination susceptibility and scorer validation.
Agentic SLM Ensembles: How Tokenomics Beat GPT-5.4 by 5.81% on IFEval
Agentic SLM ensembles exceed GPT-5.4 by 5.81% on IFEval while operating in a lower-cost regime. Analyze tokenomics and operational behavior.
ArchitectureIQ Benchmark: LLMs Struggle with Model Training Intuition
LLMs achieve 76% accuracy on ArchitectureIQ vs. human researchers’ 66%. Discover the limitations of LLM training intuition.
Detecting Hallucination in LLMs: Tracing the Topological Signatures of Impaired Context Sharing
Analyze attention graph topology to detect hallucinations in LLMs using Forman-Ricci curvature and impaired context sharing patterns.