Line of inquiry
Inquiring lines›How should we train models for cap…›What systematic failures and vulne…›this line of inquiry
Does fine-tuning modify underlying model capabilities or only behavioral outputs?
A broader line of inquiry — a family of 44 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 44
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does fine-tuning actually change model capabilities or only output distribution?
- How do finetuning and pretraining improvements differ in their effects on model capabilities?
- What happens to base model capabilities when you apply finetuning?
- Does pretraining data size matter less than base model scale for finetuning?
- Why do proprietary models improve with training while open-source models decline?
- How does behavioral fine-tuning differ from factual knowledge encoding in models?
- Does fine-tuning a small model match fine-tuning a large one?
- Why does mixed instruction data sometimes hurt specific model capabilities?
- Why does fine-tuning change how models process retrieved context?
- How does model scale affect anticipatory behavior in structured training?
- What capabilities actually require massive scale versus specialized training regimes?
- Why do smaller and larger models converge on different output formats?
- What mechanisms cause overly hard samples to degrade prior model performance?
- Why does the gap between theoretical expressiveness and learned capability matter?
- Can models adapt and combine search strategies beyond their training algorithm?
- Why does fine-tuning fail to remove temporal contamination from pretraining?
- Does importance sampling actually recover capabilities lost to hard sample training?
- What hidden costs emerge when you fine-tune models for a single domain?
- Why do sparse parameter subsets enable full-rank learning in RL?
- How does the functional separation of knowledge and reasoning affect adaptation methods?
- How much performance is lost when converting pretrained checkpoints versus training from scratch?
- Why does adaptation concentrate in low-dimensional subspaces of weights or representations?
- Can expert vectors learned offline transfer across multiple model architectures?
- Which finetuning method works best across different task and data regimes?
- How does pretrained knowledge constrain what adaptation strategies can achieve?
- What happens to model capability as weight sparsity increases during training?
- Can finetuning sparse subnetworks alone match full parameter finetuning results?
- What trade-offs emerge between training objectives and model reliability?
- How does distributional distance from pre-training relate to model difficulty?
- Do newer language model generations improve forecasting ability without additional training?
- Why does the order of training examples matter for what models learn?
- Do different domains require different types of model investment?
- Why does parameter-efficient tuning scaling fail to improve finetuning performance?
- Why does training order matter across different domain types?
- Does preference tuning help or hurt the exploration of solution spaces in code?
- What makes utility-weighted training backfire in machine learning systems?
- Can ensemble predictions be distilled back into a single deployable model?
- Can we predict out-of-distribution generalization without access to downstream tasks?
- How should guidance levels adapt as the model's capability boundary shifts?
- Can specialized components replace single fully-trained models in deployment?
- Can training data analysis predict which samples will cause unintended personality changes?
- What access constraints allow description-based adaptation but block conventional techniques?
- How much does pretraining quality affect the modularity of fine-tuned models?
- What makes two timescales better than one for minimizing weight movement?