BWhat changed
Integrated chunked audio encoding and downsampled audio-language adaptation for long-audio conversations, released open weights and speech understanding/reasoning evaluation resources, and documented transcription-versus-understanding training tradeoffs.
Scientific scope
Integrated chunked audio encoding and downsampled audio-language adaptation for long-audio conversations, released open weights and speech understanding/reasoning evaluation resources, and documented transcription-versus-understanding training tradeoffs. Whisper encoder, chunk attention, DPO and language-model backbones retain prior credit. Model durations depend on token budgets. Original Small release kept SFT because preference tuning degraded ASR; do not claim every released variant uses online DPO. Later speech synthesis is a separate contribution.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Whisper encoder, chunk attention, DPO and language-model backbones retain prior credit. Model durations depend on token budgets. Original Small release kept SFT because preference tuning degraded ASR; do not claim every released variant uses online DPO. Later speech synthesis is a separate contribution.
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.
- Voxtral
- Mistral AI: Voxtral long-audio understanding and evaluation resource — official institutional account
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