arXiv:2512.16245v2 Announce Type: replace Abstract: Merging large language models (LLMs) is a practical way to compose capabilities from multiple fine-tuned checkpoints without retraining. Yet standa...
arXiv:2510. 17426v3 Announce Type: replace-cross Abstract: The "alignment tax" of post-training is typically framed as a drop in task accuracy.
arXiv:2610. 34855v1 Announce Type: new Abstract: Visual counting benchmarks report whether vision-language models produce the correct number, but not whether errors arise from perception or reasoning.
The paper "Stress-testing Alignment Midtraining" examines the effectiveness of alignment midtraining (AMT), a technique that continues pretraining on alignment-relevant data to improve generalisation beyond post‑training methods. Experiments on models up to 110 billion parameters and 1 billion midtraining tokens reveal that AMT can steer a model’s motivation in simple scenarios, but its effects are quickly overridden by even a tiny fraction of finetuning data with a competing motivation. The study also shows that rule-following requires demonstrations in either the midtraining or post‑training datasets to be robustly learned, leading the authors to conclude that current public evidence is insufficient to confirm that AMT resolves the core alignment challenges of powerful AI systems.
The study investigates whether language models tailored to specific cognitive domains better align with corresponding brain systems. By prompting and fine‑tuning large language models into six domain experts—sensory, spatial, numerical, reasoning, social, and abstract—the authors find that each expert’s representations more closely match the brain region associated with its domain than other experts. This domain‑specific alignment holds across multiple base models and fMRI datasets, while overall prediction accuracy remains largely unchanged, indicating that regional alignment can be obscured when summarizing across the brain.
arXiv:2610.33495v1 Announce Type: new Abstract: Combining language models is usually done by averaging weights, which works only when the models share a common starting point. We ask whether five fro...
arXiv:2606. 24716v1 Announce Type: cross Abstract: Sparse autoencoders (SAEs) are increasingly used to extract interpretable concepts from vision and vision language models, yet existing evaluation methods largely rely on proxy metrics or qualitative inspection rather than measuring semantic correspondence.
arXiv:2610.06896v1 Announce Type: new Abstract: Frontier multimodal large language models (MLLMs) are increasingly positioned as general purpose visual reasoners as part of the quest for artificial g...
arXiv:2606. 19549v1 Announce Type: new Abstract: Low-rank adaptation (LoRA) makes it cheap to train many domain- and task-specific language model adapters, but whether two adapters can be merged is usually discovered only after both have been fully trained and evaluated.
arXiv:2606. 26387v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) extend large language models (LLMs) with visual perception, enabling joint reasoning over images and text.
arXiv:2412. 01282v2 Announce Type: replace-cross Abstract: Vision-Language Models (VLMs) bring powerful understanding and reasoning capabilities to multimodal tasks.
The paper introduces the task alignment proxy, a method that accelerates hyperparameter selection for merging models fine‑tuned on diverse vision tasks. It addresses the challenge of training heterogeneous decoders, which makes traditional downstream performance evaluation costly. By using the proxy, the authors demonstrate that model merging can be applied efficiently to multi‑task vision models beyond CLIP‑based classification.