Stateful streaming ASR
Beyond Fresh Starts: Stateful Inference for Streaming ASR in Conversational Voice Agents
Sameep Chattopadhyay, Alexander Erdmann, Mari Ostendorf · 2026-08-22
- Modality
- Audio & speech
- Data
- Real data
- Task · ground truth
- Gen./State · Text+WER
- Scope
- component evaluation
What it tests
- CECross-modal evidenceNot evaluated
- LELong-range evidencePartially evaluated
- SCState changesPartially evaluated
- CFConflicting memoriesNot evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionNot evaluated
- SPSpeaker trackingPartially evaluated
- ACAcoustic-cue retentionPartially evaluated
- RPRepair persistenceNot evaluated
Survey notes
Cache carryover and rollback are assessed through ASR accuracy on conversational speech. This is a speech-component state protocol rather than an agent-memory benchmark.