TL;DR of the paper, generated automatically.

Speculative Listening guesses the next audio chunk while idle, cutting response delays.

• How: drafts the next chunk assuming silence, then verifies it when audio arrives.
• Speed: speak chunks 2.6-2.8x faster, tool calls up to 23.4x.
• Latency: response delay down 73%, near the 200 ms human turn gap.