SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.26693 · 2026

Model Merging on Loss Landscape: A Geometry Perspective

Brian Axelrod, Tristan Emrich, Juanwu Lu, Anand Bhaskar, Ekaterina Tolstaya

arXiv · PDF · Open in the Atlas

Code that ran

4 of 8 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 4 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2605.26693.json · badge: /api/ran/2605.26693.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
mlfoundations/task_vectors canonical (classifier label)
classifier basis: census
4 ran · 4 no recorded run
FunctionStatusWhere it lives
convert Ran
ran; our contract draft was wrong, not the code
mlfoundations/task_vectors/src/datasets/cifar10.py
pointer only (licence: NONE); the stored copy get_code serves, from BAIKLAB/SiM, is pointer only (licence: NONE) · get_code("6d77ee840479c381")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

cosine_lr Ran
ran; our contract draft was wrong, not the code
mlfoundations/task_vectors/src/utils.py
pointer only (licence: NONE); the stored copy get_code serves, from BAIKLAB/SiM, is pointer only (licence: NONE) · get_code("2d5d7a12d8de02b7")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

maybe_dictionarize Ran
ran; our contract draft was wrong, not the code
mlfoundations/task_vectors/src/datasets/common.py
pointer only (licence: NONE) · get_code("b1e9e0d3d7d7e615")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

pretify_classname Ran
ran; our contract draft was wrong, not the code
mlfoundations/task_vectors/src/datasets/eurosat.py
pointer only (licence: NONE) · get_code("241a2a00ba0c307b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

accuracy Not yet run
unverified (harvested, no recorded run)
mlfoundations/task_vectors/src/utils.py
pointer only (licence: NONE) · get_code("aafcff3ad1ccef6e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_features Not yet run
unverified (harvested, no recorded run)
mlfoundations/task_vectors/src/datasets/common.py
pointer only (licence: NONE) · get_code("3b97c12751c0cf5b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_features_helper Not yet run
unverified (harvested, no recorded run)
mlfoundations/task_vectors/src/datasets/common.py
pointer only (licence: NONE) · get_code("238ff32fcbf410d6")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

torch_load_old Not yet run
unverified (harvested, no recorded run)
mlfoundations/task_vectors/src/utils.py
pointer only (licence: NONE); the stored copy get_code serves, from BAIKLAB/SiM, is pointer only (licence: NONE) · get_code("7d051a7adefedb6c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, mlfoundations/task_vectors when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 8 are also present in other papers whose code we hold (8 with at least one paper dated earlier, 3 the same month, 4 later) and 0 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
accuracy 22 other papers
20 dated earlier, 1 the same month, 1 later
2207.05333 (2022-07)
IDEA: Increasing Text Diversity via Online Multi-Label Recognition for Vision-Language Pre-training
2208.05592 (2022-08)
Patching open-vocabulary models by interpolating weights
2211.12914 (2022-11)
Open-vocabulary Attribute Detection
and 19 more not named here
maybe_dictionarize 13 other papers
12 dated earlier, 1 the same month · 1 only as an assembly brick
2308.01313 (2023-08)
PerceptionCLIP: Visual Classification by Inferring and Conditioning on Contexts
2308.12213 (2023-08)
CLIPN for Zero-Shot OOD Detection: Teaching CLIP to Say No
2311.01723 (2023-11)
Towards Calibrated Robust Fine-Tuning of Vision-Language Models
and 10 more not named here
cosine_lr 11 other papers
9 dated earlier, 1 the same month, 1 later
2208.05592 (2022-08)
Patching open-vocabulary models by interpolating weights
2305.12827 (2023-05)
Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models
2308.12213 (2023-08)
CLIPN for Zero-Shot OOD Detection: Teaching CLIP to Say No
and 8 more not named here
convert 8 other papers
7 dated earlier, 1 later
2208.05592 (2022-08)
Patching open-vocabulary models by interpolating weights
2312.15901 (2023-12)
Black-Box Tuning of Vision-Language Models with Effective Gradient Approximation
2405.07813 (2024-05)
Localizing Task Information for Improved Model Merging and Compression
and 5 more not named here
torch_load_old 7 other papers
6 dated earlier, 1 later
2305.12827 (2023-05)
Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained Models
2410.05583 (2024-10)
NegMerge: Sign-Consensual Weight Merging for Machine Unlearning
2410.14389 (2024-10)
SurgeryV2: Bridging the Gap Between Model Merging and Multi-Task Learning with Deep Representation Surgery
and 4 more not named here
get_features_helper 6 other papers
6 dated earlier
2405.07813 (2024-05)
Localizing Task Information for Improved Model Merging and Compression
2406.05184 (2024-06)
The Unmet Promise of Synthetic Training Images: Using Retrieved Real Images Performs Better
2410.17146 (2024-10)
LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging
and 3 more not named here
get_features 5 other papers
5 dated earlier
2405.07813 (2024-05)
Localizing Task Information for Improved Model Merging and Compression
2410.17146 (2024-10)
LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging
2411.06966 (2024-11)
Robust Fine-tuning of Zero-shot Models via Variance Reduction
and 2 more not named here
pretify_classname 5 other papers
5 dated earlier
2304.04512 (2023-04)
Defense-Prefix for Preventing Typographic Attacks on CLIP
2405.07813 (2024-05)
Localizing Task Information for Improved Model Merging and Compression
2410.17146 (2024-10)
LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging
and 2 more not named here

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Model merging offers a promising avenue to knowledge integration and parallel development without retraining. Yet, existing methods either ignore the geometry of the loss landscape or rely on intractable full-space Hessian approximations. We propose EpiMer, a framework that casts model merging as solving the Fréchet mean on a Riemannian manifold and restricts the computation to a low-rank subspace spanned by the task vectors. With the expected Hessian as the metric, we reveal a connection between local curvature and epistemic uncertainty of the parameters. Our theoretical analysis decomposes the merging error bound into the subspace Fréchet variance and the residual energy, and provides a closed-form characterization of when curvature-aware merging provably outperforms flatgeometry methods. In addition, our framework unifies both curvature-aware methods and recent spectral methods as special cases of the subspace Fréchet mean with different geometric metrics. Merging fine-tuned CLIP-ViT models on eight image classification tasks, Epistemic Merging strictly outperforms the baselines on all three CLIP-ViT backbones at matched rank, improving the across-task average accuracy and worst-task accuracy on every backbone.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2605.26693")
get_code_for_paper("2605.26693")
have("2605.26693")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2605.26693.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2605.26693.svg)](https://syntology.ai/paper/2605.26693)

Connect an agent — have() is free.