BWhat changed
Used a decoder-only multistream model over pre-aligned streams at a shared frame rate, introducing relative delays to trade latency against quality for streaming speech recognition and synthesis.
Scientific scope
Used a decoder-only multistream model over pre-aligned streams at a shared frame rate, introducing relative delays to trade latency against quality for streaming speech recognition and synthesis. Inherits Moshi multistream architecture and wait-k ideas. Alignment is moved to preprocessing, not eliminated; finite-context processing supports continuing streams without guaranteeing long-range memory or uniformly accurate unbounded sequences.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits Moshi multistream architecture and wait-k ideas. Alignment is moved to preprocessing, not eliminated; finite-context processing supports continuing streams without guaranteeing long-range memory or uniformly accurate unbounded sequences.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Streaming Sequence-to-Sequence Learning with Delayed Streams Modeling
- Kyutai scientific publications index
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction