# Preserve to Expand: Source-Aware Adaptation and Model Merging for Multilingual LLM-Based ASR

Source: [arXiv](https://arxiv.org/abs/2610.35846v1)  
Feed7 permalink: https://feed7.dev/p/2610-35846v1-0krwatc  
Published: 2026-10-26T17:56:22.000Z  
Trust: Needs Review (needs_review)

## Why Included

A study of multilingual speech-model adaptation finds a trade-off between learning new languages and preserving existing ones. Update merging and merge order can change how well the original languages are retained.

## Source Summary

The paper studies adapting a trainable speech connector while keeping the language model frozen. In single-target expansion, **SSU-Wanda+TIES** gave the best reported balance of learning a new language and retaining source-language performance; target-only tuning and standard TIES caused severe forgetting.

## Practical Implication

For speech-model builders adding languages without replaying source-language data, compare both target adaptation and source retention. In multi-target expansion, update merging retained source performance more consistently, while recursive expansion could improve target adaptation depending on language order.

## Agent-Ready Context

The paper studies adapting a trainable speech connector while keeping the language model frozen. In single-target expansion, **SSU-Wanda+TIES** gave the best reported balance of learning a new language and retaining source-language performance; target-only tuning and standard TIES caused severe forgetting.

For speech-model builders adding languages without replaying source-language data, compare both target adaptation and source retention. In multi-target expansion, update merging retained source performance more consistently, while recursive expansion could improve target adaptation depending on language order.

The findings concern connector-based Speech LLMs. The material does not provide enough detail to generalize the results to other model architectures or training setups.

## Connected Context

Feed7 judgment across 928 accumulated Signals:

This supplies a concrete, connector-specific example of multilingual expansion where adaptation quality depends on retaining source-language ability as well as learning the target. The results favor particular merge strategies in the tested Speech LLM setup, so they sharpen adaptation choices without establishing a general recipe for other architectures.

- [Adaption Labs: Gradient-Free Continual Learning — Sara Hooker, Adaption](https://feed7.dev/p/adaption-labs-gradient-free-continual-learning-sara-hooker-adaption-1pozjjc) — Both concern adapting models while preserving useful existing capability; this paper reports specific connector-merging and retention outcomes, whereas Auto Scientist describes broader automated search over adaptation recipes without comparable evidence here.
- [Program-as-Weights: A Programming Paradigm for Fuzzy Functions](https://feed7.dev/p/2607-02512v1-1dr5458) — Both adapt a frozen language-model backbone through trainable added components, but the target studies speech connectors and language retention while Program-as-Weights addresses fuzzy-function execution.

## Context Map

- Layer: model
- Domains: audio, research
- Topics: open-models, model-selection

## Uncertainty

- The findings concern connector-based Speech LLMs. The material does not provide enough detail to generalize the results to other model architectures or training setups.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.