pith. sign in

arxiv: 2310.12808 · v2 · pith:TFTMXUYRnew · submitted 2023-10-19 · 💻 cs.LG · cs.AI· cs.CL

Model Merging by Uncertainty-Based Gradient Matching

classification 💻 cs.LG cs.AIcs.CL
keywords averagingheremodelsperformanceuncertainty-basedweighted-averagingarithmeticassumptions
0
0 comments X
read the original abstract

Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail? Here, we connect the inaccuracy of weighted-averaging to mismatches in the gradients and propose a new uncertainty-based scheme to improve the performance by reducing the mismatch. The connection also reveals implicit assumptions in other schemes such as averaging, task arithmetic, and Fisher-weighted averaging. Our new method gives consistent improvements for large language models and vision transformers, both in terms of performance and robustness to hyperparameters. Code available here.

This paper has not been read by Pith yet.

discussion (0)

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

Forward citations

Cited by 5 Pith papers

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

  1. Model Merging as Probabilistic Inference in Fine-Tuning Parameter Space

    cs.LG 2026-07 unverdicted novelty 7.0

    Model merging is cast as PoE inference with EBM experts, revealing Gaussian assumptions in prior work and proposing convergent Cauchy experts that improve empirical performance.

  2. 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.

  3. Unlocking the Potential of Continual Model Merging: An ODE Perspective

    cs.LG 2026-05 unverdicted novelty 6.0

    Introduces ODE-M, an ODE-based merging method for continual model merging that follows low-loss connecting paths to mitigate catastrophic forgetting.

  4. Unlocking the Potential of Continual Model Merging: An ODE Perspective

    cs.LG 2026-05 unverdicted novelty 6.0

    ODE-M traces low-loss connecting paths via time-dependent velocity fields and barrier constraints to improve controllability and reduce forgetting in continual model merging.

  5. Unlocking the Potential of Continual Model Merging: An ODE Perspective

    cs.LG 2026-05 unverdicted novelty 6.0

    ODE-M formulates continual model merging as a barrier-aware ODE trajectory in parameter space, using first-order feedback and a utility-aware schedule to balance retained knowledge and new task performance.