model merging
- Papers
- 13
- Released code
- 0
- First seen
- Oct 2026
- Latest
- Nov 2026
13 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1Language models can merge specialists without matching their training capacity
Model merging does not require uniform LoRA ranks; per-domain rank choice boosts averaging methods but hurts magnitude-based conflict resolution.
Nov 2026 - 2Image restoration models can share skills without full joint retraining
OptIMerge merges image-to-image diffusion models without retraining by learning structured weight combinations from few examples.
Nov 2026 - 3Merging AI models improves by removing conflicting feature directions
Joint Null-Space Projection merges fine-tuned models without data, achieving 89.45% accuracy on eight CLIP tasks by removing cross-task subspace interference.
Nov 2026
How fast it is moving
Often studied with
reinforcement learning with verifiable rewardsReinforcement learning that uses rewards from objectively checkable signals like correct answers or verification feedback.2 togetherbackdoor attackA hidden trigger in a model causes malicious behavior only when activated, while the model otherwise appears benign.2 togethercontinual learningA learning paradigm where a model sequentially acquires knowledge from multiple tasks without significantly forgetting earlier ones.1 togetherThe papers
Most central to this idea first, not most recent.
- Academiccs.CV
Merging AI models improves by removing conflicting feature directions
Joint Null-Space Projection merges fine-tuned models without data, achieving 89.45% accuracy on eight CLIP tasks by removing cross-task subspace interference.
Tohoku University · Nov 2026
- Big Techcs.LG
Language models can merge specialists without matching their training capacity
Model merging does not require uniform LoRA ranks; per-domain rank choice boosts averaging methods but hurts magnitude-based conflict resolution.
Amazon, Amazon Web Services · Nov 2026
- Big Techcs.CV
Image restoration models can share skills without full joint retraining
OptIMerge merges image-to-image diffusion models without retraining by learning structured weight combinations from few examples.
Google · Nov 2026
- Chinese Techcs.AI
Merging reasoning models layer by layer cuts answers by up to 93%
Low-Rank Hierarchical Merging outperforms prior methods on long-to-short reasoning model merging by using second-order curvature and dual constraints.
University of Virginia, University of California, Los Angeles · Nov 2026
- Chinese Techcs.CR
Dormant backdoor can lie hidden in one model and activate on merging
A backdoor can be hidden in a single language model so it only activates when the model is merged with others, causing safety failures.
Zhejiang University, Alibaba Group · Oct 2026
- Chinese Techcs.LG
Distilled task updates can merge better than stronger teacher updates
On-policy distillation task vectors can complement RL teacher updates and compose more effectively across tasks.
Institute of Automation, Chinese Academy of Sciences, University of Chinese Academy of Sciences · Oct 2026
- Chinese Techcs.AI
Small language models judge outputs more accurately when trained on counterfactual pairs
Counterfactual pairwise augmentation (CPA) improves pointwise judges by training on label-flipping response pairs.
Huawei Technologies Ltd. · Nov 2026
- Top Universitycs.LG
A single attack triggers malicious behavior in LLMs across multiple deployment-time transformations
An attack creates LLMs that appear benign but become malicious when quantization, pruning, or merging is applied.
ETH Zurich · Oct 2026
- Independentcs.LG
RL training runs can yield stronger policies than their best checkpoints
An RL training history can yield policies stronger than its best checkpoint without extra training or inference compute.
Oct 2026
- Chinese Techcs.LG
Averaging checkpoints across shorter training runs beats one long mid-training run
Distributing mid-training compute across multiple independent training trajectories improves LLM performance more than a single long run.
Ant Group, Shanghai Jiao Tong University · Oct 2026
- Academiccs.CV
AI image classifiers can share combined updates with little accuracy loss
Summing same-task LoRAs from different pretrained backbones works without alignment, nearly matching solo performance.
Chiba University, National Institute of Informatics · Oct 2026
- Top Universitycs.AI
A simple diagnostic predicts whether two AI models will work well together before costly combination
Model pair compatibility can be predicted with local neighborhood overlap and paradigm ceiling before alignment or merging.
EPFL, University of Basel · Oct 2026
- Top Universitycs.LG
Random Outputs Can Replace Model Updates in Collaborative Learning
Knowledge can be shared between models by transmitting only outputs on random, task-unrelated inputs.
Università della Svizzera italiana (USI), MIT Media Lab · Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.