相关论文: Making Models Unmergeable via Scaling-Sensitive Lo…

Model merging aims to cheaply combine individual task-specific models into a single multitask model. In this work, we view past merging methods as leveraging different notions of a ''task parameter subspace'' in which models are matched…

Increasing model size has unlocked a dazzling array of capabilities in modern language models. At the same time, even frontier models remain vulnerable to jailbreaks and prompt injections, despite concerted efforts to make them robust. As…

The "alignment tax" of post-training is typically framed as a drop in task accuracy. We show it also involves a severe loss of calibration, making models overconfident, less reliable, and model outputs less diverse. We show that this…

Artificial intelligence is retracing the Internet's path from centralized provision to distributed creation. Initially, resource-intensive computation concentrates within institutions capable of training and serving large models.Eventually,…

Model merging dramatically reduces storage and computational resources by combining multiple expert models into a single multi-task model. Although recent model merging methods have shown promising results, they struggle to maintain…

Recent work on permutation-based model merging has shown impressive low- or zero-barrier mode connectivity between models from completely different initializations. However, this line of work has not yet extended to the Transformer…

Model merging for Large Language Models (LLMs) directly fuses the parameters of different models finetuned on various tasks, creating a unified model for multi-domain tasks. However, due to potential vulnerabilities in models available on…

Merging multiple expert models offers a promising approach for performing multi-task learning without accessing their original data. Existing methods attempt to alleviate task conflicts by sparsifying task vectors or promoting orthogonality…

Fine-tuning-as-a-service, while commercially successful for Large Language Model (LLM) providers, exposes models to harmful fine-tuning attacks. As a widely explored defense paradigm against such attacks, unlearning attempts to remove…

Model merging combines the parameters of multiple neural networks into a single model without additional training. As fine-tuned large language models (LLMs) proliferate, merging offers a computationally efficient alternative to ensembles…

Recent advancements in building domain-specific large language models (LLMs) have shown remarkable success, especially in tasks requiring reasoning abilities like logical inference over complex relationships and multi-step problem solving.…

Model merging aims to efficiently combine the weights of multiple expert models, each trained on a specific task, into a single multi-task model, with strong performance across all tasks. When applied to all but the last layer of weights,…

The availability of deep hedging has opened new horizons for solving hedging problems under a large variety of realistic market conditions. At the same time, any model - be it a traditional stochastic model or a market generator - is at…

Modern text-to-image (T2I) models amplify harmful societal biases, challenging their ethical deployment. We introduce an inference-time method that reliably mitigates social bias while keeping prompt semantics and visual context…

The quality of open-weight language models has dramatically improved in recent years. Sharing weights greatly facilitates model adoption by enabling their use across diverse hardware and software platforms. They also allow for more open…

Gradient sparsification, while mitigating communication bottlenecks in Federated Learning (FL), fundamentally alters the geometric landscape of model updates. We reveal that the resultant high-dimensional orthogonality renders traditional…

Model merging has shown great promise at combining expert models, but the benefit of merging is unclear when merging "generalist" models trained on many tasks. We explore merging in the context of large (~100B) models, by recycling…

Watermarking has become a practical tool for tracing language model outputs, but it modifies token probabilities at inference time, which were carefully tuned by alignment training. This creates a tension: how do watermark-induced shifts…

Model merging combines fine-tuned checkpoints into a single multi-task model without retraining. Existing methods - such as task arithmetic, model soups, TIES, and DARE - are computationally efficient and empirically successful, but rely on…

Open-sourcing foundation models (FMs) enables broad reuse but also exposes model trainers to economic and safety risks from unrestricted downstream fine-tuning. We address this problem by building non-fine-tunable foundation models: models…