mlsanityData to 7 Oct 2026

Deep Model Merging: The Sister of Neural Network Interpretability -- A Survey

arXiv 2410.12927 ↗First posted 16 Oct 2024cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

We survey the model merging literature through the lens of loss landscape geometry to connect observations from empirical studies on model merging and loss landscape analysis to phenomena that govern neural network training and the emergence of their inner representations. We distill repeated empirical observations from the literature in these fields into descriptions of four major characteristics of loss landscape geometry: mode convexity, determinism, directedness, and connectivity. We argue that insights into the structure of learned representations from model merging have applications to model interpretability and robustness, subsequently we propose promising new research directions at the intersection of these fields.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 94 of its 137 references that are in the map. First posted on arXiv 16 Oct 2024.

Builds on

94 of 137 references.

References · 94

  1. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  4. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  5. 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
  6. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  7. 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,413
  8. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
  9. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  10. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  11. 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,157
  12. 2022Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference timecited by 1,366
38 more
  1. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  2. 2016On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minimacited by 1,337
  3. 2018Averaging Weights Leads to Wider Optima and Better Generalizationcited by 1,194
  4. 2017Visualizing the Loss Landscape of Neural Netscited by 1,126
  5. 2022Mass-Editing Memory in a Transformercited by 976
  6. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
  7. 2016Practical Black-Box Attacks against Machine Learningcited by 893
  8. 2022Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learningcited by 886
  9. 2023TIES-Merging: Resolving Interference When Merging Modelscited by 795
  10. 2015Optimizing Neural Networks with Kronecker-factored Approximate Curvaturecited by 776
  11. 2024DoRA: Weight-Decomposed Low-Rank Adaptationcited by 740
  12. 2023Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunchcited by 679
  13. 2020Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuningcited by 658
  14. 2019Linear Mode Connectivity and the Lottery Ticket Hypothesiscited by 627
  15. 2021Merging Models with Fisher-Weighted Averagingcited by 613
  16. 2018Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNscited by 585
  17. 2020Federated Learning with Matched Averagingcited by 564
  18. 2017Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Datacited by 527
  19. 2020Dataset Condensation with Gradient Matchingcited by 510
  20. 2020Pruning neural networks without any data by iteratively conserving synaptic flowcited by 475
  21. 2022Git Re-Basin: Merging Models modulo Permutation Symmetriescited by 460
  22. 2019Deep Ensembles: A Loss Landscape Perspectivecited by 438
  23. 2021Gradient Descent on Neural Networks Typically Occurs at the Edge of Stabilitycited by 413
  24. 2019Bayesian Nonparametric Federated Learning of Neural Networkscited by 385
  25. 2022Dataless Knowledge Fusion by Merging Weights of Language Modelscited by 384
  26. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 381
  27. 2019A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protectioncited by 371
  28. 2024LoRA+: Efficient Low Rank Adaptation of Large Modelscited by 365
  29. 2018Measuring the Intrinsic Dimension of Objective Landscapescited by 364
  30. 2016Entropy-SGD: Biasing Gradient Descent Into Wide Valleyscited by 357
  31. 2023A Comprehensive Survey on Test-Time Adaptation under Distribution Shiftscited by 353
  32. 2017Snapshot Ensembles: Train 1, get M for freecited by 349
  33. 2018Essentially No Barriers in Neural Network Energy Landscapecited by 328
  34. 2021The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networkscited by 285
  35. 2024Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunitiescited by 247
  36. 2019Uniform convergence may be unable to explain generalization in deep learningcited by 223
  37. 2019Model Fusion via Optimal Transportcited by 217
  38. 2024Evolutionary Optimization of Model Merging Recipescited by 204

50 most cited of 94 All 94 (JSON)

Built on by

Not cited yet.