---
title: "Will it Merge? On The Causes of Model Mergeability (2026)"
description: "The authors define a measurable \"mergeability score.\" A model update (a LoRA adapter) is repeatedly merged… 99% accuracy. 32 references."
authors: ["Adir Rahamim", "Asaf Yehudai", "Boaz Carmeli", "Leshem Choshen", "Yosi Mass", "Yonatan Belinkov"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.1322/"
doi: "10.18653/v1/2026.findings-acl.1322"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/rahamim-2026-will-it-merge-causes-model/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/rahamim-2026-will-it-merge-causes-model/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/48/) / Will it Merge? On The Causes of Model Mergeability

# Will it Merge? On The Causes of Model Mergeability

**Authors:** Adir Rahamim, [Asaf Yehudai](https://md.research.elamur.ai/free/en/authors/asaf-yehudai/), Boaz Carmeli, [Leshem Choshen](https://md.research.elamur.ai/free/en/authors/leshem-choshen/), [Yosi Mass](https://md.research.elamur.ai/free/en/authors/yosi-mass/) and [Yonatan Belinkov](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.1322/) · [PDF](https://aclanthology.org/2026.findings-acl.1322.pdf) · [DOI 10.18653/v1/2026.findings-acl.1322](https://doi.org/10.18653/v1/2026.findings-acl.1322) · [OpenAlex](https://openalex.org/W7166808471)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Merging combines several fine-tuned models into one multitask model without retraining, yet it is unclear why some fine-tuned updates survive merging better than others. The authors define a measurable "mergeability score." A model update (a LoRA adapter) is repeatedly merged with random subsets of other adapters, and the score is its average performance on its own data across these trials. They test this at two levels. At the example level, they use PopQA with Llama-3.2-3B and Qwen-2.5-3B. At the task level, they use the Lots-of-LoRAs collection on Mistral-7B-Instruct-v0.2, merging mainly with KnOTS. They then compare three candidate explanations: the base model's knowledge of the specific task, weight properties, and general domain knowledge. The strongest predictor is how well the base model already handles the fine-tuning data. Updates targeting knowledge the base model nearly has merge more robustly, while weakly known tasks degrade most. Mergeability also behaves as a local property of each update and is largely independent of its merge partners. As a proof of concept, the authors propose weighting the merge by base-model performance, which better preserves weak tasks.

## Key points
- **PopQA setup:**
  - Questions were converted to 8-option multiple choice (7 distractors) and evaluated with 4-shot prompting and length-normalized log-probabilities.
  - Out of 1931 questions, 1107 were already answered correctly by the base model, which left 824 to fine-tune.
  - Of those 824, 639 (77.55%) were answered correctly after per-example LoRA training on Wikipedia passages about the entity.
- **Lots-of-LoRAs setup:** only tasks whose adapter reached at least 99% accuracy were kept, giving 81 tasks. Accuracy was measured by exact match.
- **Example-level predictor:** a small probability gap in the base model between its top answer and the gold answer correlates with higher mergeability.
- **Rejected explanations:** structural weight properties and general domain knowledge correlate poorly with mergeability.
- **Locality of the trait:** highly mergeable adapters stay stable regardless of which mergeability group they are merged with. The main patterns largely hold for other merging algorithms (Appendix A.9).

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Massachusetts Institute of Technology](https://md.research.elamur.ai/free/en/universities/massachusetts-institute-of-technology/) — University, Cambridge, United States · 211 works · 69,000 citations
- [IBM](https://md.research.elamur.ai/free/en/companies/ibm/) — Company, Armonk, United States · 139 works · 70,691 citations
- [Harvard University](https://md.research.elamur.ai/free/en/universities/harvard-university/) — University, Cambridge, United States · 112 works · 42,023 citations
- [Hebrew University of Jerusalem](https://md.research.elamur.ai/free/en/universities/hebrew-university-of-jerusalem/) — University, Jerusalem, Israel · 65 works · 9,387 citations
- [Technion – Israel Institute of Technology](https://md.research.elamur.ai/free/en/universities/technion-israel-institute-of-technology/) — University, Haifa, Israel · 49 works · 5,306 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works

## Cite

Cite as: Adir Rahamim, Asaf Yehudai, Boaz Carmeli, Leshem Choshen, Yosi Mass and Yonatan Belinkov. 2026. Will it Merge? On The Causes of Model Mergeability. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{rahamim2026will,
  title = {{Will it Merge? On The Causes of Model Mergeability}},
  author = {Adir Rahamim, Asaf Yehudai, Boaz Carmeli, Leshem Choshen, Yosi Mass and Yonatan Belinkov},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1322},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1322}
}
```

- [← previous WildSci: Advancing Scientific Reasoning from In-the-Wild Literature](https://md.research.elamur.ai/free/en/papers/liu-2026-wildsci-advancing-scientific-reasoning-wild/)
- [next → WinoQueer-NL: Assessing Bias in Dutch Language Models toward LGBTQ+ Identities](https://md.research.elamur.ai/free/en/papers/beuk-2026-winoqueer-nl-assessing-bias-dutch/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/48/)

## References

32 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Samuel K Ainsworth, Jonathan Hayase, and Siddhartha Srinivasa. 2023. Git re-basin: Merging models modulo permutation symmetries. In The Eleventh International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Samuel+K+Ainsworth%2C+Jonathan+Hayase%2C+and+Siddhartha+Srinivasa%2E+2023%2E+Git+re%2Dbasin%3A+Merging+models+modulo+permutation+symmetries%2E+In+The+Eleventh+International+Conference+on+Learning+Representations%2E)
2. Viraat Aryabumi, Yixuan Su, Raymond Ma, Adrien Morisot, Ivan Zhang, Acyr Locatelli, Marzieh Fadaee, Ahmet Üstün, and Sara Hooker. 2024. To code or not to code? exploring impact of code in pre-training. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Viraat+Aryabumi%2C+Yixuan+Su%2C+Raymond+Ma%2C+Adrien+Morisot%2C+Ivan+Zhang%2C+Acyr+Locatelli%2C+Marzieh+Fadaee%2C+Ahmet+%C3%9Cst%C3%BCn%2C+and+Sara+Hooker%2E+2024%2E+To+code+or+not+to+code%3F+exploring+impact+of+code+in+pre%2Dtraining)
3. Sameer Bansal, H. Kamper, Karen Livescu, Adam Lopez, and S. Goldwater. 2018. [Pre-training on high-resource speech recognition improves low-resource speech-to-text translation](https://md.research.elamur.ai/free/en/papers/bansal-2019-pre-training-high-resource-speech/). In North American Chapter of the Association for Computational Linguistics.
4. Joachim Bingel and Anders Søgaard. 2017. [Identifying beneficial task relations for multi-task learning in deep neural networks](https://md.research.elamur.ai/free/en/papers/bingel-2017-identifying-beneficial-task-relations-multi/). In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 164–169, Valencia, Spain. Asso…
5. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. [Language models are few-shot learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/). Advances in neural information processing systems, 33:1877–1901.
6. Rickard Brüel-Gabrielsson, Jiacheng Zhu, Onkar Bhardwaj, Leshem Choshen, Kristjan Greenewald, Mikhail Yurochkin, and Justin Solomon. 2025. Compress then serve: Serving thousands of lora adapters with little overhead. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Rickard+Br%C3%BCel%2DGabrielsson%2C+Jiacheng+Zhu%2C+Onkar+Bhardwaj%2C+Leshem+Choshen%2C+Kristjan+Greenewald%2C+Mikhail+Yurochkin%2C+and+Justin+Solomon%2E+2025%2E+Compress+then+serve%3A+Serving+thousands+of+lora+adapters+with+)
7. Leshem Choshen, Elad Venezian, Shachar Don-Yehiya, Noam Slonim, and Yoav Katz. 2023. Where to start? analyzing the potential value of intermediate models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 1446–1470, Singapore. Association for Computatio… [find](https://scholar.google.com/scholar?q=Leshem+Choshen%2C+Elad+Venezian%2C+Shachar+Don%2DYehiya%2C+Noam+Slonim%2C+and+Yoav+Katz%2E+2023%2E+Where+to+start%3F+analyzing+the+potential+value+of+intermediate+models%2E+In+Proceedings+of+the+2023+Conference+on+Empi)
8. Leshem Choshen, Elad Venezian, Noam Slonim, and Yoav Katz. 2022. Fusing finetuned models for better pretraining. arXiv preprint arXiv:2204.03044. — [arXiv:2204.03044](https://arxiv.org/abs/2204.03044) [find](https://scholar.google.com/scholar?q=Leshem+Choshen%2C+Elad+Venezian%2C+Noam+Slonim%2C+and+Yoav+Katz%2E+2022%2E+Fusing+finetuned+models+for+better+pretraining%2E+arXiv+preprint+arXiv%3A2204%2E03044%2E+%E2%80%94+%5BarXiv%3A2204%2E03044%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2204%2E03044%29)
9. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,…
10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv e-prints, pages arXiv–2407.
11. Almog Gueta, Elad Venezian, Colin Raffel, Noam Slonim, Yoav Katz, and Leshem Choshen. 2023. Knowledge is a region in weight space for fine-tuned language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1350–1370, Singapore. Association for Computational Lingui… [find](https://scholar.google.com/scholar?q=Almog+Gueta%2C+Elad+Venezian%2C+Colin+Raffel%2C+Noam+Slonim%2C+Yoav+Katz%2C+and+Leshem+Choshen%2E+2023%2E+Knowledge+is+a+region+in+weight+space+for+fine%2Dtuned+language+models%2E+In+Findings+of+the+Association+for+Com)
12. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). arXiv preprint arXiv:2501.12948. 26559 9
13. Guy Hacohen and Daphna Weinshall. 2019. On the power of curriculum learning in training deep networks. In International conference on machine learning, pages 2535–2544. PMLR. [find](https://scholar.google.com/scholar?q=Guy+Hacohen+and+Daphna+Weinshall%2E+2019%2E+On+the+power+of+curriculum+learning+in+training+deep+networks%2E+In+International+conference+on+machine+learning%2C+pages+2535%E2%80%932544%2E+PMLR%2E)
14. Yifei He, Siqi Zeng, Yuzheng Hu, Rui Yang, Tong Zhang, and Han Zhao. 2025. Mergebench: A benchmark for merging domain-specialized llms. arXiv preprint arXiv:2505.10833. — [arXiv:2505.10833](https://arxiv.org/abs/2505.10833) [find](https://scholar.google.com/scholar?q=Yifei+He%2C+Siqi+Zeng%2C+Yuzheng+Hu%2C+Rui+Yang%2C+Tong+Zhang%2C+and+Han+Zhao%2E+2025%2E+Mergebench%3A+A+benchmark+for+merging+domain%2Dspecialized+llms%2E+arXiv+preprint+arXiv%3A2505%2E10833%2E+%E2%80%94+%5BarXiv%3A2505%2E10833%5D%28https%3A%2F%2Far)
15. Stefan Horoi, Guy Wolf, Eugene Belilovsky, and Gintare Karolina Dziugaite. 2025. Less is more: Undertraining experts improves model upcycling. arXiv preprint arXiv:2506.14126. — [arXiv:2506.14126](https://arxiv.org/abs/2506.14126) [find](https://scholar.google.com/scholar?q=Stefan+Horoi%2C+Guy+Wolf%2C+Eugene+Belilovsky%2C+and+Gintare+Karolina+Dziugaite%2E+2025%2E+Less+is+more%3A+Undertraining+experts+improves+model+upcycling%2E+arXiv+preprint+arXiv%3A2506%2E14126%2E+%E2%80%94+%5BarXiv%3A2506%2E14126%5D%28htt)
16. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). ICLR, 1(2):3.
17. Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Keming Lu, and 1 others. 2024. Qwen2. 5-coder technical report. arXiv preprint arXiv:2409.12186. — [Qwen2.5-Coder Technical Report](https://md.research.elamur.ai/free/en/papers/hui-2024-qwen2-5-coder-technical-report/)
18. Gabriel Ilharco, Marco Tulio Ribeiro, Mitchell Wortsman, Suchin Gururangan, Ludwig Schmidt, Hannaneh Hajishirzi, and Ali Farhadi. 2023. [Editing models with task arithmetic](https://md.research.elamur.ai/free/en/papers/ilharco-2022-editing-models-task-arithmetic/). In The Eleventh International Conference on Learning Representations.
19. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… [find](https://scholar.google.com/scholar?q=Albert+Q%2E+Jiang%2C+Alexandre+Sablayrolles%2C+Arthur+Mensch%2C+Chris+Bamford%2C+Devendra+Singh+Chaplot%2C+Diego+de+las+Casas%2C+Florian+Bressand%2C+Gianna+Lengyel%2C+Guillaume+Lample%2C+Lucile+Saulnier%2C+L%C3%A9lio+Renard+Lav)
20. Keller Jordan, Hanie Sedghi, Olga Saukh, Rahim Entezari, and Behnam Neyshabur. 2023. Repair: Renormalizing permuted activations for interpolation repair. In ICLR. [find](https://scholar.google.com/scholar?q=Keller+Jordan%2C+Hanie+Sedghi%2C+Olga+Saukh%2C+Rahim+Entezari%2C+and+Behnam+Neyshabur%2E+2023%2E+Repair%3A+Renormalizing+permuted+activations+for+interpolation+repair%2E+In+ICLR%2E)
21. Joongwon Kim, Akari Asai, Gabriel Ilharco, and Hannaneh Hajishirzi. 2023. TaskWeb: Selecting better source tasks for multi-task NLP. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 11032–11052, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Joongwon+Kim%2C+Akari+Asai%2C+Gabriel+Ilharco%2C+and+Hannaneh+Hajishirzi%2E+2023%2E+TaskWeb%3A+Selecting+better+source+tasks+for+multi%2Dtask+NLP%2E+In+Proceedings+of+the+2023+Conference+on+Empirical+Methods+in+Natur)
22. Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. When not to trust language models: Investigating effectiveness of parametric and non-parametric memories. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (V… [find](https://scholar.google.com/scholar?q=Alex+Mallen%2C+Akari+Asai%2C+Victor+Zhong%2C+Rajarshi+Das%2C+Daniel+Khashabi%2C+and+Hannaneh+Hajishirzi%2E+2023%2E+When+not+to+trust+language+models%3A+Investigating+effectiveness+of+parametric+and+non%2Dparametric+mem)
23. Jyothish Pari, Samy Jelassi, and Pulkit Agrawal. 2025. Collective model intelligence requires compatible specialization. In ICLR 2025 Workshop on Modularity for Collaborative, Decentralized, and Continual Deep Learning. [find](https://scholar.google.com/scholar?q=Jyothish+Pari%2C+Samy+Jelassi%2C+and+Pulkit+Agrawal%2E+2025%2E+Collective+model+intelligence+requires+compatible+specialization%2E+In+ICLR+2025+Workshop+on+Modularity+for+Collaborative%2C+Decentralized%2C+and+Conti)
24. Clifton Poth, Jonas Pfeiffer, Andreas Rücklé, and Iryna Gurevych. 2021. [What to pre-train on? Efficient intermediate task selection](https://md.research.elamur.ai/free/en/papers/poth-2021-what-pre-train-efficient-intermediate/). In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 10585–10605, Online and Punta Cana, Dominican Republic. Association fo…
25. Abhinav Shrivastava, Abhinav Gupta, and Ross Girshick. 2016. Training region-based object detectors with online hard example mining. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 761–769. [find](https://scholar.google.com/scholar?q=Abhinav+Shrivastava%2C+Abhinav+Gupta%2C+and+Ross+Girshick%2E+2016%2E+Training+region%2Dbased+object+detectors+with+online+hard+example+mining%2E+In+Proceedings+of+the+IEEE+conference+on+computer+vision+and+patter)
26. George Stoica, Pratik Ramesh, Boglarka Ecsedi, Leshem Choshen, and Judy Hoffman. 2025. Model merging with svd to tie the knots. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=George+Stoica%2C+Pratik+Ramesh%2C+Boglarka+Ecsedi%2C+Leshem+Choshen%2C+and+Judy+Hoffman%2E+2025%2E+Model+merging+with+svd+to+tie+the+knots%2E+In+The+Thirteenth+International+Conference+on+Learning+Representations%2E)
27. Mitchell Wortsman, Gabriel Ilharco, Samir Ya Gadre, Rebecca Roelofs, Raphael Gontijo-Lopes, Ari S Morcos, Hongseok Namkoong, Ali Farhadi, Yair Carmon, Simon Kornblith, and 1 others. 2022. [Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time](https://md.research.elamur.ai/free/en/papers/wortsman-2022-model-soups-averaging-weights-multiple/)…
28. Prateek Yadav, Derek Tam, Leshem Choshen, Colin A Raffel, and Mohit Bansal. 2023. [Ties-merging: Resolving interference when merging models](https://md.research.elamur.ai/free/en/papers/yadav-2023-ties-merging-resolving-interference-merging/). Advances in Neural Information Processing Systems, 36:7093–7115.
29. Prateek Yadav, Tu Vu, Jonathan Lai, Alexandra Chronopoulou, Manaal Faruqui, Mohit Bansal, and Tsendsuren Munkhdalai. 2025. What matters for model merging at scale? Transactions on Machine Learning Research. [find](https://scholar.google.com/scholar?q=Prateek+Yadav%2C+Tu+Vu%2C+Jonathan+Lai%2C+Alexandra+Chronopoulou%2C+Manaal+Faruqui%2C+Mohit+Bansal%2C+and+Tsendsuren+Munkhdalai%2E+2025%2E+What+matters+for+model+merging+at+scale%3F+Transactions+on+Machine+Learning+Res)
30. Le Yu, Bowen Yu, Haiyang Yu, Fei Huang, and Yongbin Li. 2024. [Language models are super mario: Absorbing abilities from homologous models as a free lunch](https://md.research.elamur.ai/free/en/papers/yu-2023-language-models-super-mario-absorbing/). In Forty-first International Conference on Machine Learning.
31. Kerem Zaman, Leshem Choshen, and Shashank Srivastava. 2024. Fuse to forget: Bias reduction and selective memorization through model fusion. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 18763–18783. [find](https://scholar.google.com/scholar?q=Kerem+Zaman%2C+Leshem+Choshen%2C+and+Shashank+Srivastava%2E+2024%2E+Fuse+to+forget%3A+Bias+reduction+and+selective+memorization+through+model+fusion%2E+In+Proceedings+of+the+2024+Conference+on+Empirical+Methods+i)
32. Michael S Matena and Colin A Raffel. 2022. [Merging models with fisher-weighted averaging](https://md.research.elamur.ai/free/en/papers/matena-2021-merging-models-fisher-weighted-averaging/). Advances in Neural Information Processing Systems, 35:17703– 17716. 26560 10

*Updated 1 November 2026*