paper-with-me

홈 › Papers

AlignMerge - Alignment-Preserving Large Language Model Merging via Fisher-Guided Geometric Constraints

2025-12-18 · Aniruddha Roy, Jyoti Patel, Aman Chadha, Vinija Jain, Amitava Das arxiv

Merging large language models (LLMs) is a practical way to compose capabilities from multiple fine-tuned checkpoints without retraining. Yet standard schemes (linear weight soups, task vectors, and Fisher-weighted averaging) can preserve loss while quietly destroying alignment. We argue that merging is not a numerical trick but a geometry-constrained operation around an already-aligned anchor: fusion must be steered to respect safety geometry, not validated post hoc. We introduce AlignMerge, a geometry-aware merging framework that makes alignment an explicit invariant. In a local Fisher chart around an instruction-tuned base, we estimate an alignment subspace with projector P_A and optimize: L_AlignMerge = L_geo + lambda_align * L_align + lambda_bud * L_bud, where L_geo keeps the merge close to its experts in Fisher-Rao geometry, L_align penalizes motion along alignment-sensitive directions, and L_bud enforces a soft alignment budget. As the alignment functional we use the decoding-invariant Alignment Quality Index (AQI), a latent-space criterion that captures how cleanly aligned and misaligned behaviors separate in representation space. Across five model families (LLaMA-3 8B, Mistral 7B, Qwen 2, Phi-3.5, Gemma 2), merging safety anchors with task experts, AlignMerge improves alignment metrics (AQI, toxicity, LLM-judge alignment) while matching or exceeding the best expert on instruction-following, reasoning, and helpfulness. It also exhibits smaller alignment-subspace drift and fewer budget violations than Fisher soups, TIES, SafeMerge, and MergeAlign. These results make alignment-preserving merging a first-class design goal and suggest a path to geometry-aware composition of future foundation models.

📄 PDF Abstract BibTeX arXiv:2512.16245

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging

2025-03-21 · Aladin Djuhera, Swanand Ravindra Kadhe, Farhan Ahmed, Syed Zawad 외

Fine-tuning large language models (LLMs) on downstream tasks can inadvertently erode their safety alignment, even for benign fine-tuning datasets. We address this challenge by proposing SafeMERGE, a post-fine-tuning fram…

GSM8KSafety Alignment

Pruning via Merging: Compressing LLMs via Manifold Alignment Based Layer Merging

2024-06-24 · Deyuan Liu, Zhanyue Qin, Hairu Wang, Zhao Yang 외

While large language models (LLMs) excel in many domains, their complexity and scale challenge deployment in resource-limited environments. Current compression techniques, such as parameter pruning, often fail to effecti…

MMLUModel CompressionQuantization

Accurate and Efficient Low-Rank Model Merging in Core Space

2025-09-22 · Aniello Panariello, Daniel Marczak, Simone Magistri, Angelo Porrello 외 arxiv

In this paper, we address the challenges associated with merging low-rank adaptations of large neural networks. With the rise of parameter-efficient adaptation techniques, such as Low-Rank Adaptation (LoRA), model fine-t…

Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning

2025-12-10 · Lama Alssum, Hani Itani, Hasan Abed Al Kader Hammoud, Philip Torr 외 arxiv

The safety alignment of large language models (LLMs) is becoming increasingly important with their democratization. In this paper, we study the safety degradation that comes with adapting LLMs to new tasks. We attribute …

Continual Learning

A Data-Efficient Path to Multilingual LLMs: Language Expansion via Post-training PARAM$Δ$ Integration into Upcycled MoE

2026-05-18 · Hao Zhou, Tianhao Li, Zhijun Wang, Shuaijie She 외 arxiv

Expanding Large Language Models~(LLMs) to new languages is a costly endeavor, demanding extensive Continued Pre-Training~(CPT) and data-intensive alignment. While recent data-free merging techniques attempt to bypass ali…

Language Acquisition