Speech-model backdoor fires on one speaker’s voice
In sandbox tests, a poisoned transcript asking for $5,000 led some GPT-4o and Claude agents to pay.
64
Product
Whisper
- Mentions · 7d1
- Mentions · 30d3
- All mentions4
- First seenSep 22, 2026
- Last mentioned3d ago
Timeline
4 articles · newest firstOpenPA: Forced-alignment-free Multi-granular Pronunciation Assessment with Mixed-transcript Adaptation
OpenPA uses Whisper cross-attention to score pronunciation at word and utterance level without forced alignment.
26
Speech event detector: Timestamped detection of non-verbal vocalizations and background sounds for rich speech transcription
A 5.8M model timestamps 54 non-speech sounds and merges them into Whisper transcripts.
42
Six Layers Less: Encoder Pruning for Whisper with Label-Free Recovery
Six Whisper encoder layers are pruned and partly recovered with label-free distillation.
44
Appears with
- OpenAIvendor · 3
- Whisperai model · 2
- NVIDIAvendor · 1
- Anthropicvendor · 1
- Claude Opusai model · 1
- Claude Haikuai model · 1
- whisper-large-v3-turbo-encoder-prunedai model · 1
- whisper-large-v3-turboai model · 1
- Imperial College Londonorganization · 1
- Qwenai model · 1
- GPT-4oai model · 1
- Parakeetai model · 1
- Whisper-mediumai model · 1
- Speech Event Detectorai model · 1
- Claude Sonnetai model · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.