ACL Findings2026
Skill Weaving: Efficient LLM Improvement via Modular Skillpacks
Zhuo Li, Guodong Du, Zesheng Shi, Weiyang Guo, Weijun Yao, Yuan Zhou, Jiabo Zhang, Jing Li
Abstract
Large language models increasingly require specialization across diverse domains, yet existing approaches struggle to balance multidomain capacities with strict memory and inference constraints. In this work, we introduce SkillWeave, a modular improvement framework that enables LLMs to specialize under fixed memory budgets. SkillWeave partitions full capabilities of a general-purpose model into skillpacks-lightweight, domainspecific delta modules-that reorganize and refine the model's internal knowledge. For efficient deployment, SkillWeave integrates Skil-lZip to compress skillpacks into compact and inference-ready format, enabling strong multidomain performance with low-latency execution. On multi-task and agentic benchmarks, a 9B SkillWeave model outperforms several baselines and even surpasses a 32B monolithic LLM, while achieving up to 4× speedup.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback cites Direct Preference Optimization: Your Language Model is Secretly a Reward Model
- SimPO: Simple Preference Optimization with a Reference-Free Reward cites Direct Preference Optimization: Your Language Model is Secretly a Reward Model
Builds on28
- Direct Preference Optimization: Your Language Model is Secretly a Reward ModelRafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning et al.NeurIPS 2023 · 10,924 citations
- SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language ModelsGuangxuan Xiao, Ji Lin, Mickaël Seznec, Hao Wu et al.ICML 2023 · 1,493 citations
- SimPO: Simple Preference Optimization with a Reference-Free RewardYu Meng, Mengzhou Xia, Danqi ChenNeurIPS 2024 · 1,203 citations
- Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free LunchLe Yu, Bowen Yu, Haiyang Yu, Fei Huang et al.ICML 2024 · 605 citations
- RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI FeedbackHarrison Lee, Samrat Phatale, Hassan Mansoor, Thomas Mesnard et al.ICML 2024 · 598 citations
Related papers
- TrimLLM: Progressive Layer Dropping for Domain-Specific LLMsLanxiang Hu, Tajana Rosing, Hao ZhangACL 2025
- SALAAD: Sparse And Low-Rank Adaptation via ADMM for Large Language Model InferenceHao Ma, Melis Ilayda Bal, Liang Zhang, Bingcong Li et al.ICML 2026
- ModularMoE: Fast LLM Customization with Parameter-Sharing Mixture-of-Experts for Low-Resource SettingsJiaxing Liu, Qi Qi, Haifeng Sun, Dunjun Li et al.ACL Findings 2026
- A Learnable Skill Combination Strategy for Multi-task Learning in Natural Language UnderstandingZhe Yang, Yi Huang, Yaqin Chen, Mengfei Guo et al.ACL Findings 2026
- Synergy over Discrepancy: A Partition-Based Approach to Multi-Domain LLM Fine-TuningHua Ye, Siyuan Chen, Haoliang Zhang, Weihao Luo et al.NeurIPS 2025 · 2 citations
