Speech decisions without decoding: turn-taking, emotion, non-verbal sounds and more from raw speech, in one forward pass.
-
Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript
Paper • 2610.02638 • Published -
DuplexJev — speech to decisions
🎧10 voice-agent decisions from raw speech in one pass
-
DuplexJev Vision — voice + picture
👁Judge a voice against a picture in one forward pass
-
adventists-ai/DuplexJev-32B-Turn
Audio-Text-to-Text • 33B • Updated • 26