REVIEW 7 cited by
Merging Models with Fisher-Weighted Averaging
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Averaging the parameters of models that have the same architecture and initialization can provide a means of combining their respective capabilities. In this paper, we take the perspective that this "merging" operation can be seen as choosing parameters that approximately maximize the joint likelihood of the posteriors of the models' parameters. Computing a simple average of the models' parameters therefore corresponds to making an isotropic Gaussian approximation to their posteriors. We develop an alternative merging procedure based on the Laplace approximation where we approximate each model's posterior as a Gaussian distribution whose precision matrix corresponds to its Fisher information. We first show that our "Fisher merging" technique provides a performance boost in settings where simple parameter averaging is currently used -- specifically, robust fine-tuning and model ensembling. Then, we compare merging to standard gradient-based transfer learning and demonstrate that merging enables a fundamentally different method for transferring capabilities across models. Specifically, we show that Fisher merging is competitive with gradient-based transfer learning approaches (while being significantly cheaper) in intermediate-task training and domain-adaptive pre-training. We also show that our merging procedure makes it possible to combine models in previously unexplored ways. We release our code to facilitate future research into methods for merging models.
Forward citations
Cited by 7 Pith papers
-
Unlearning as Distribution Restoration: A Controlled Counterfactual Study, a Validated Selective Screen, and the Limits of Oracle-Free Certification
Matching a retrained oracle on trained probes can certify models that still retain held-out forget knowledge, and oracle-free unlearning certification is only possible for counterfactual, non-inferable facts.
-
Forget the Data and Fine-Tuning! Just Fold the Network to Compress
Model folding compresses a network by k-means clustering similar neurons across adjacent layers and repairing activation statistics without data (Fold-AR, Fold-DIR), surpassing prior data-free methods at high sparsity.
-
Rethinking Heterogeneous LLM Merging: A Weighted Model Averaging Perspective
After truncating or expanding checkpoints to a shared shape, small-ratio weight averaging slightly improves average benchmark scores over strong Qwen sources, but headline gains are inflated by per-task best-ratio selection.
-
AlphaWiSE: Adaptive Weight Interpolation for Continual Multimodal Representation Learning
Fitting one interpolation coefficient per parameter tensor on a small exemplar memory improves continual audio–image–text retrieval over individual continual-learning checkpoints.
-
DeltaMerge-LowRes: Composing Language and Task Deltas for Low-Resource Adaptation
A TIES-style merge across language-delta and task-delta axes (cross-axis TIES) beats additive and task-only composition on low-resource summarisation and QA.
-
Exploring Sparse Adapters for Scalable Merging of Parameter Efficient Experts
Sparse adapters trained with max connection sensitivity outperform LoRA and full fine-tuning both alone and after merging 20 task experts, but still lag multitask training on unseen tasks.
-
Revisiting Bayesian Model Averaging in the Era of Foundation Models
The paper proposes Bayesian model averaging and an entropy-minimizing weight optimizer for ensembling foundation models, reporting accuracy gains over output averaging on image and text classification tasks.
Discussion (0). Sign in to comment.