EchoChain: A Full-Duplex Benchmark for State-Update Reasoning Under Interruptions

Create a testing toolkit for voice agents to evaluate behavior under interruption. It fills a massive gap in current agent benchmarking where focus is only on full response completion.

Suggested repo: echo-bench

"Don't let interruptions break your voice assistant's logic."

Estimated effort: 45h