Crimson AI NewsA CrimsonLingua Network service
EN ع
← Back to news
Research paper Hugging Face

Voice Memory: A Training-Free Approach to Agentic Speech Recognition

AI By Crimson AI Hugging Face Papers 30 July 2026 · 00:00 16 views
Share: X Telegram

Hugging Face researchers introduce Voice Memory, an inference-only scheme that couples a frozen ASR decoder with a frozen corrector via a per-domain memory file, cutting weighted WER from 8.36% to 7.52% across ten domains with zero parameter updates.

Voice Memory: A Training-Free Approach to Agentic Speech Recognition

Key points

Researchers at Hugging Face have unveiled Voice Memory, a novel inference-only scheme for agentic speech recognition that requires no weight updates or added parameters. The system introduces a listener-thinker architecture, where a frozen ASR decoder (the listener) and a frozen corrector (the thinker) interact solely through a per-domain memory file (memory.md).

At stream time, the corrector reads the memory and decides per utterance whether to act on the hypothesis or abstain, keeping the 1-best result. Asynchronously, a score-gated optimizer revises the memory through bounded edits, accepting changes only when they strictly improve a held-out score. This design keeps the learned skill auditable and portable, as no model weights are altered.

A key finding is that restraint is the operative skill the loop discovers. Unconstrained generative error correction (GER) over-corrects, breaking correct tokens on up to 64% of its edits on financial news. Voice Memory reduces this harmful-edit rate to 35%.

Across ten HyPoradise domains with an open corrector, Voice Memory lowers weighted word error rate (WER) from 8.36% to 7.52% (7.47% with three added in-context examples) without regressing any dataset below its 1-best baseline. Gains are most pronounced in air-travel commands (8.40% to 3.40%) and noisy far-field speech (CHiME-4, 12.69% to 10.46%).

The memory transfers across corrector families and adds zero parameters to the inference path. The deployed WSJ memory is just 776 bytes. A demo and example code are available for future studies.

MetricBaseline (1-best)Voice MemoryVoice Memory + 3 in-context examples
Weighted WER (10 HyPoradise domains)8.36%7.52%7.47%
Air-travel commands WER8.40%3.40%-
CHiME-4 WER12.69%10.46%-
Harmful-edit rate (financial news)64%35%-
Source
Hugging Face · Hugging Face Papers
Related news
Research paper
Hugging Face 31 Aug 2026

Hugging Face Unveils StepGuard: Step-Level Guardrails for Safer AI Agents

StepGuard, a new step-level guard model from Hugging Face, audits agent actions before execution, reducing attack success rates by...

1
Research paper
Hugging Face 31 Aug 2026

Hugging Face Researchers Unveil ABot-Recon for Stable Long-Horizon 3D Reconstruction

ABot-Recon, a new streaming 3D reconstruction model from Hugging Face, achieves stable long-horizon performance using only local t...

1
Research paper
Hugging Face 31 Aug 2026

ContextPilot: Teaching Agents Proactive Context Management via Fine-Grained RL

Hugging Face researchers introduce ContextPilot, a framework that enhances long-horizon agent reasoning by expanding context-editi...

1