SPIN Unprocessed July 30, 2026 ai_technology research
Voice Memory for Agentic Speech Recognition
View original on arxiv.orgOverview
arXiv:2607.26410v1 Announce Type: new Abstract: We present Voice Memory, a inference-only scheme for agentic speech recognition: at stream time, a frozen corrector reads a single per-domain memory.md and decides per utterance whether to act on the hypothesis or abstain and keep the 1-best. Asynchronously, a score-gated optimizer revises that file through bounded edits, accepting an edit only when it strictly improves a held-out score. Extended from classical ASR-LM framework, we refer this split
SpinGraph analysis pending — check back after processing.
Ask AI about this story
Opens with the SpinGraph .md URL and structured context — one click, prompt included.
More from arXiv Computation and Language
View all →- CMT-RAG: Complementary Memory Traces for Multi-turn Multi-hop RAG
- ForgetBench: Benchmarking Forgetting Dynamics of Long-Term Parametric Memory in Language Models
- Mergeable Model-Side Aggregation States for Long-Context Language Models
- Knowledge before Reasoning: EC-Reason-Bench, a Training-Free Diagnostic Benchmark for LLM Enzyme Classification
- Misalignment Has a Personality: A Big Five Account of Emergent Misalignment
- (Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding
Markdown (.md) · JSON-LD schema (.json) · Machine-readable for AI & GEO