In the news
Preserve to Expand: Source-Aware Adaptation and Model Merging for Multilingual LLM-Based ASR
arXiv cs.AI · Published · 3 min read
In 30 seconds
- What happened
- Researchers evaluated model merging strategies for expanding multilingual speech recognition systems to new languages while preserving performance on previously supported languages.
- Why it matters
- Engineers building multilingual ASR systems need this when adding new languages without access to original training data for existing supported languages.
- Watch out
- Multi-target expansion is highly order-sensitive; recursive expansion improves target adaptation only under favorable language orderings, limiting practical applicability.
- llm
- language model
- fine-tun
- serving
- speech
The patterns behind this
- MAPS: Multilingual Agent Performance & Security
- Agent Context Preservation and Recovery
- Agentic Context Engineering (Evolving Playbook)
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.