Section

Language Models

5 pieces · Artificial Intelligence, Machine Learning & Computing Research · Est. 2007

Language Models

Benchmark Contamination Is Breaking Model Evaluation: What a Credible Pipeline Looks Like

Benchmark contamination makes MMLU and HumanEval unreliable for frontier models. What a credible, contamination-resistant evaluation pipeline requires.

By Institute for Joint Cognition & AI · 9 min · 26 June 2026

Language Models

Agentic System Failures Are Orchestration Problems, Not Model Problems

Multi-agent deployments fail mostly on orchestration, not weak models — race conditions, memory gaps, timeouts. Evaluation must look beyond task success.

By Institute for Joint Cognition & AI · 10 min · 26 June 2026

Language Models

Evaluating LLMs: What Benchmarks Measure and Where They Fail

A technical guide to LLM evaluation: what MMLU, HELM, GSM8K, and HumanEval actually measure, benchmark contamination, and how to build reliable task-specific evals.

By Institute for Joint Cognition & AI · 13 min · 10 June 2024

Language Models

Long-Context Retrieval: What Actually Works When the Context Window Isn't Enough

Why bigger context windows don't fix retrieval quality, what 'lost in the middle' means in practice, and when to use RAG, long context, or both.

By Institute for Joint Cognition & AI · 12 min · 25 April 2024

Language Models

Retrieval-Augmented Generation in Practice: Building RAG That Works

A technical guide to retrieval augmented generation: chunking, embeddings, vector search, reranking, failure modes, and how to evaluate your RAG pipeline.

By Institute for Joint Cognition & AI · 13 min · 22 January 2024