speech recognition
Speech recognition is technology that turns spoken words into written text, like when you dictate a message or talk to a voice assistant. Modern systems use deep learning models that learn patterns in sound and language, but they often need powerful computers or cloud servers to run. Researchers are working to make these models fast, accurate, and usable on cheap devices, for different languages and accents, and honest about when they might be wrong.
The papers here show that speech recognition is moving in several directions at once: making it work on tiny microcontrollers, testing it across diverse dialects and languages, and correcting its overconfidence in garbled audio. They also reveal that current models still struggle with real-world variation, dialects, noise, and demographic differences, and that improving fairness and error awareness is as important as cutting word error rates. Together, the work pushes toward more capable, efficient, and trustworthy voice interfaces.
- Papers
- 8
- Released code
- 0
- First seen
- Sept 2026
- Latest
- Oct 2026
8 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1Pruned CTC cuts memory for large-vocabulary ASR training without sacrificing accuracy.
Pruned CTC restricts CTC alignment computation to target tokens and blank, making large-vocabulary ASR training memory-efficient with exactly equivalent loss and gradients.
Oct 2026 - 2New benchmark measures speech recognition across 17 Arabic dialects
GPT-4o-transcribe achieves 35% WER on 17 Arabic dialects, but performance varies considerably across groups.
Oct 2026 - 3Open-vocabulary speech recognition runs on a $5 microcontroller without a neural accelerator
Oído brings open-vocabulary English speech recognition to a $5 microcontroller without a neural accelerator.
Oct 2026
How fast it is moving
Often studied with
large language modelA neural network trained on massive text corpora to generate coherent text and perform diverse language tasks.5 togetherbenchmarkA standard set of tasks or data used to evaluate and compare the performance of AI systems.2 togetherconfidence calibrationConfidence calibration aligns a model's predicted confidence with its actual accuracy across predictions.1 togetherWho is working on it
The papers
Most central to this idea first, not most recent.
- Industrycs.SD
Open-vocabulary speech recognition runs on a $5 microcontroller without a neural accelerator
Oído brings open-vocabulary English speech recognition to a $5 microcontroller without a neural accelerator.
Lokutor · Oct 2026
- Chinese Techeess.AS
Pruned CTC cuts memory for large-vocabulary ASR training without sacrificing accuracy.
Pruned CTC restricts CTC alignment computation to target tokens and blank, making large-vocabulary ASR training memory-efficient with exactly equivalent loss and gradients.
Shanghai Jiao Tong University, University of Cambridge · Oct 2026
- Independentcs.CL
New benchmark measures speech recognition across 17 Arabic dialects
GPT-4o-transcribe achieves 35% WER on 17 Arabic dialects, but performance varies considerably across groups.
Oct 2026
- Big Tech
Audio LLMs cannot assess their own transcription reliability.
Audio-encoder representations predict whether an Audio LLM will transcribe degraded speech reliably far better than self-assessment or tested quality and uncertainty baselines.
Apple, University of California San Diego · Sept 2026
- Big Techcs.CL
Language discrimination training reduces the multilingual gap in speech models
Strengthening language discrimination during pretraining closes the multilingual gap in speech models.
Apple · Oct 2026
- Big Techcs.CL
Open voice model combines real-time conversation with native tool calling
NemotronLabs VoiceChat is a full-duplex speech-to-speech model that natively invokes tools during conversation.
NVIDIA · Sept 2026
- Independentcs.AI
Voice-agent models diverge most beyond choosing the correct tool
Most voice agent failures come from argument values and rule compliance, not from selecting the wrong tool.
Sept 2026
- Independentcs.SD
Real-voice benchmark reveals model-specific demographic bias in Audio-LLMs
HEAR is a 87k-sample real-voice, demographically diverse benchmark showing voice bias in Audio-LLMs is model-specific and worsened by personalization.
Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.