Pith. sign in

REVIEW 10 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

arxiv 2111.09832 v2 pith:DNCMRHB5 submitted 2021-11-18 cs.LG

Merging Models with Fisher-Weighted Averaging

classification cs.LG
keywords mergingmodelsparametersaveragingfisherapproximationcapabilitiescorresponds
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
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.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 10 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Editing Models with Task Arithmetic

    cs.LG 2022-12 accept novelty 8.0

    Task vectors from weight differences allow arithmetic operations to edit pre-trained models, improving multiple tasks simultaneously and enabling analogical inference on unseen tasks.

  2. Unlearning as Distribution Restoration: A Controlled Counterfactual Study, a Validated Selective Screen, and the Limits of Oracle-Free Certification

    cs.LG 2026-07 conditional novelty 7.0

    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.

  3. Task Prompt Vectors: Effective Initialization through Multi-Task Soft-Prompt Transfer

    cs.CL 2024-08 unverdicted novelty 7.0

    Task prompt vectors, formed by subtracting random initialization from tuned soft prompts, support low-resource initialization and arithmetic combination across tasks on 12 NLU datasets while remaining independent of i...

  4. Quantifying the Agreement Between Data-Influence and Data-Similarity to Understand LLM Behavior

    cs.LG 2026-06 unverdicted novelty 6.0

    Data-similarity and data-influence produce significantly overlapping rankings of training documents for LLM outputs, with asymmetry allowing a favorable cost-accuracy trade-off.

  5. Spectral Unforgetting: Post-Hoc Recovery of Damaged Capabilities Without Retraining

    cs.LG 2026-05 unverdicted novelty 6.0

    DG-Hard uses Donoho-Gavish hard thresholding on the fine-tuning weight delta to separate task-aligned signal from noise-like residual, recovering damaged capabilities while preserving target-task gains.

  6. Experience Sharing in Mutual Reinforcement Learning for Heterogeneous Language Models

    cs.LG 2026-05 unverdicted novelty 6.0

    Mutual Reinforcement Learning allows heterogeneous LLMs to exchange experience through mechanisms like Peer Rollout Pooling, Cross-Policy GRPO Advantage Sharing, and Success-Gated Transfer, with outcome-level sharing ...

  7. Routing-Based Continual Learning for Multimodal Large Language Models

    cs.LG 2025-11 unverdicted novelty 6.0

    Routing architecture for MLLMs enables continual learning with constant compute, matching multi-task learning performance and supporting cross-modal transfer.

  8. Rethinking Heterogeneous LLM Merging: A Weighted Model Averaging Perspective

    cs.AI 2026-07 conditional novelty 5.0

    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.

  9. AlphaWiSE: Adaptive Weight Interpolation for Continual Multimodal Representation Learning

    cs.CV 2026-07 conditional novelty 5.0

    Fitting one interpolation coefficient per parameter tensor on a small exemplar memory improves continual audio–image–text retrieval over individual continual-learning checkpoints.

  10. DeltaMerge-LowRes: Composing Language and Task Deltas for Low-Resource Adaptation

    cs.CL 2026-07 conditional novelty 5.0

    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.