Research

Paper

TESTING March 19, 2026

MOSAIC: Multi-Objective Slice-Aware Iterative Curation for Alignment

Authors

Yipu Dou, Wang Yang

Abstract

We study how to allocate a fixed supervised fine-tuning budget when three objectives must be balanced at once: multi-turn safety alignment, low over-refusal on benign boundary queries, and instruction following under verifiable constraints. We propose MOSAIC (Multi-Objective Slice-Aware Iterative Curation for Alignment), a multi-objective framework for closed-loop data mixture search built on a unified L1-L3 evaluation interface. MOSAIC turns slice-level failure profiles into executable data actions, including dataset-level mixture ratios, bucket-level weights, and focus criteria. Under a fixed 1M-token budget and five rounds of independent fine-tuning from the same base model, MOSAIC improves internal XGuard from 2.76 to 4.67 while keeping OrBench at 4.41 and IFEval at 3.65. The final Pareto solution also generalizes better than a random static LoRA baseline on independent attack, over-refusal, and capability tests, suggesting that structured failure diagnosis can serve as a practical control signal for budgeted data construction. Code is available at https://github.com/douyipu/mosaic.

Metadata

arXiv ID: 2603.18637
Provider: ARXIV
Primary Category: cs.CR
Published: 2026-03-19
Fetched: 2026-03-20 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.18637v1</id>\n    <title>MOSAIC: Multi-Objective Slice-Aware Iterative Curation for Alignment</title>\n    <updated>2026-03-19T09:00:47Z</updated>\n    <link href='https://arxiv.org/abs/2603.18637v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.18637v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>We study how to allocate a fixed supervised fine-tuning budget when three objectives must be balanced at once: multi-turn safety alignment, low over-refusal on benign boundary queries, and instruction following under verifiable constraints. We propose MOSAIC (Multi-Objective Slice-Aware Iterative Curation for Alignment), a multi-objective framework for closed-loop data mixture search built on a unified L1-L3 evaluation interface. MOSAIC turns slice-level failure profiles into executable data actions, including dataset-level mixture ratios, bucket-level weights, and focus criteria. Under a fixed 1M-token budget and five rounds of independent fine-tuning from the same base model, MOSAIC improves internal XGuard from 2.76 to 4.67 while keeping OrBench at 4.41 and IFEval at 3.65. The final Pareto solution also generalizes better than a random static LoRA baseline on independent attack, over-refusal, and capability tests, suggesting that structured failure diagnosis can serve as a practical control signal for budgeted data construction. Code is available at https://github.com/douyipu/mosaic.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CR'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-19T09:00:47Z</published>\n    <arxiv:comment>9 pages, 5 figures. Code available at https://github.com/douyipu/mosaic</arxiv:comment>\n    <arxiv:primary_category term='cs.CR'/>\n    <author>\n      <name>Yipu Dou</name>\n    </author>\n    <author>\n      <name>Wang Yang</name>\n    </author>\n  </entry>"
}