Copertina

Data-Free Layer-Adaptive Merging via Fisher Information for Long-to-Short Reasoning LLMs

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autore principale:	Xia, Tian
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

LARV: Data-Free Layer-wise Adaptive Rescaling Veneer for Model Merging
di: Wang, Xinyu, et al.
Pubblicazione: (2026)

AdaFisher: Adaptive Second Order Optimization via Fisher Information
di: Gomes, Damien Martins, et al.
Pubblicazione: (2024)

Model Merging via Data-Free Covariance Estimation
di: Hameed, Marawan Gamal Abdel, et al.
Pubblicazione: (2026)

Fishers for Free? Approximating the Fisher Information Matrix by Recycling the Squared Gradient Accumulator
di: Li, YuXin, et al.
Pubblicazione: (2025)

SyMerge: From Non-Interference to Synergistic Merging via Single-Layer Adaptation
di: Jung, Aecheon, et al.
Pubblicazione: (2024)

Data-Free Pruning of Self-Attention Layers in LLMs
di: Saikumar, Dhananjay, et al.
Pubblicazione: (2025)

Functionality-Oriented LLM Merging on the Fisher--Rao Manifold
di: Wang, Jiayu, et al.
Pubblicazione: (2026)

Fisher Mask Nodes for Language Model Merging
di: K, Thennal D, et al.
Pubblicazione: (2024)

Rethinking Layer-wise Model Merging through Chain of Merges
di: Buzzega, Pietro, et al.
Pubblicazione: (2025)

MergeMix: Optimizing Mid-Training Data Mixtures via Learnable Model Merging
di: Wang, Jiapeng, et al.
Pubblicazione: (2026)

Efficient Estimation for Longitudinal Networks via Adaptive Merging
di: Zhang, Haoran, et al.
Pubblicazione: (2022)

On the Optimality of Tracking Fisher Information in Adaptive Testing with Stochastic Binary Responses
di: Kim, Sanghwa, et al.
Pubblicazione: (2025)

Reasoning Cache: Continual Improvement Over Long Horizons via Short-Horizon RL
di: Wu, Ian, et al.
Pubblicazione: (2026)

MergeBench: A Benchmark for Merging Domain-Specialized LLMs
di: He, Yifei, et al.
Pubblicazione: (2025)

Long-Short Alignment for Effective Long-Context Modeling in LLMs
di: Du, Tianqi, et al.
Pubblicazione: (2025)

Whoever Started the Interference Should End It: Guiding Data-Free Model Merging via Task Vectors
di: Cheng, Runxi, et al.
Pubblicazione: (2025)

Adapting to Fragmented and Evolving Data: A Fisher Information Perspective
di: Khan, Behraj, et al.
Pubblicazione: (2025)

Data-light Uncertainty Set Merging with Admissibility
di: Qin, Shenghao, et al.
Pubblicazione: (2024)

Hierarchical Context Merging: Better Long Context Understanding for Pre-trained LLMs
di: Song, Woomin, et al.
Pubblicazione: (2024)

Expert Merging: Model Merging with Unsupervised Expert Alignment and Importance-Guided Layer Chunking
di: Zhang, Dengming, et al.
Pubblicazione: (2025)

Retraining-Free Merging of Sparse MoE via Hierarchical Clustering
di: Chen, I-Chun, et al.
Pubblicazione: (2024)

LayerMerge: Neural Network Depth Compression through Layer Pruning and Merging
di: Kim, Jinuk, et al.
Pubblicazione: (2024)

RAIN-Merging: A Gradient-Free Method to Enhance Instruction Following in Large Reasoning Models with Preserved Thinking Format
di: Huang, Zhehao, et al.
Pubblicazione: (2026)

Soft Merging of Experts with Adaptive Routing
di: Muqeeth, Mohammed, et al.
Pubblicazione: (2023)

QTALE: Quantization-Robust Token-Adaptive Layer Execution for LLMs
di: Noh, Kanghyun, et al.
Pubblicazione: (2026)

Adaptive Test-Time Compute Allocation for Reasoning LLMs via Constrained Policy Optimization
di: Zhai, Zhiyuan, et al.
Pubblicazione: (2026)

Scalable Model Merging with Progressive Layer-wise Distillation
di: Xu, Jing, et al.
Pubblicazione: (2025)

AutoL2S: Auto Long-Short Reasoning for Efficient Large Language Models
di: Luo, Feng, et al.
Pubblicazione: (2025)

Towards Adaptive Continual Model Merging via Manifold-Aware Expert Evolution
di: Qiu, Haiyun, et al.
Pubblicazione: (2026)

MEMIT-Merge: Addressing MEMIT's Key-Value Conflicts in Same-Subject Batch Editing for LLMs
di: Dong, Zilu, et al.
Pubblicazione: (2025)

Cross-LoRA: A Data-Free LoRA Transfer Framework across Heterogeneous LLMs
di: Xia, Feifan, et al.
Pubblicazione: (2025)

FedMerge: Federated Personalization via Model Merging
di: Chen, Shutong, et al.
Pubblicazione: (2025)

Exact Unlearning of Finetuning Data via Model Merging at Scale
di: Kuo, Kevin, et al.
Pubblicazione: (2025)

Null-Space Filtering for Data-Free Continual Model Merging: Preserving Stability, Promoting Plasticity
di: Qiu, Zihuan, et al.
Pubblicazione: (2025)

Free Energy-Driven Reinforcement Learning with Adaptive Advantage Shaping for Unsupervised Reasoning in LLMs
di: Huang, Yiming, et al.
Pubblicazione: (2026)

Encoder-Free Knowledge-Graph Reasoning with LLMs via Hyperdimensional Path Retrieval
di: Liu, Yezi, et al.
Pubblicazione: (2025)

Weight Weaving: Parameter Pooling for Data-Free Model Merging
di: Chaves, Levy, et al.
Pubblicazione: (2025)

Thinking Short and Right Over Thinking Long: Serving LLM Reasoning Efficiently and Accurately
di: Wang, Yuhang, et al.
Pubblicazione: (2025)

Demystifying Long Chain-of-Thought Reasoning in LLMs
di: Yeo, Edward, et al.
Pubblicazione: (2025)

Beyond Hidden-Layer Manipulation: Semantically-Aware Logit Interventions for Debiasing LLMs
di: Xia, Wei
Pubblicazione: (2025)