Line of inquiry
Inquiring lines›How can we ensure training objecti…›How do training approaches and fee…›this line of inquiry
How do curriculum design and feedback approaches affect model learning?
A broader line of inquiry — a family of 106 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 106
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does curriculum-based training keep small models perpetually at their learning edge?
- Can models learn better from critiquing errors than imitating correct responses?
- Does partial trace guidance work better than curriculum learning for hard problems?
- How should training incorporate external critique versus encouraging self-correction?
- Does the productive difficulty band ever stabilize during training?
- How does difficulty-adaptive curriculum learning change which samples get selected during training?
- How much can externalized skills improve models before hitting diminishing returns?
- Does training on curated solutions transfer to unseen problem types?
- Why does the gap between theoretical expressiveness and learned capability matter?
- Can model training address failures that really originate in harness gaps?
- How does training on correct answer form differ mechanistically from training on failure analysis?
- What distinguishes surface mechanisms from the training regimes that produce them?
- How do finetuning and pretraining improvements differ in their effects on model capabilities?
- How should researchers evaluate whether correct model outputs reflect real structural learning?
- How does the optimal difficulty band shift as the model's capabilities improve during training?
- What makes preventative lessons from failures more valuable than success patterns?
- Can we reverse the instruction-following deficit through targeted training?
- Why does open-ended search outperform direct curriculum building approaches?
- How does in-context learning trigger phase transitions in model behavior?
- How do different training objectives shift whether models over-predict or under-predict?
- Can trained models encode programs more complex than their data-generating process?
- How does behavioral fine-tuning differ from factual knowledge encoding in models?
- Why do weaker teacher models sometimes produce better training signals than stronger ones?
- Does the model learn depth-wise drift as an explicit strategy?
- Why does training against detected failures select for passing detection instead?
- How do complete multi-turn trajectories differ from isolated task examples?
- Why does narrow training data produce broad harmful behavior patterns?
- What training interventions could close the perception-action gap?
- What training regimes confound surface mechanisms with their actual causes?
- Why does instruction-tuning reduce a model's context-following behavior?
- Does critique training improve exploration diversity during model training or only test time?
- How do failure examples improve distillation compared to successful trajectories alone?
- Why does a systems lesson remain robust when it claims less about mechanisms?
- Why does the order of training examples matter for what models learn?
- What mechanisms cause overly hard samples to degrade prior model performance?
- Why do foundation models develop task-specific heuristics instead of general world models?
- What qualities make a behavioral pattern count as a teachable skill?
- How do transformers generate harder solutions when mostly trained on easier problems?
- Can instruction tuning succeed without explicit task understanding?
- How do difficulty metrics relate to the true value of training examples?
- What training signals would models need to learn reciprocal common-ground construction?
- What emergent behaviors do models develop when trained on underspecified pedagogical tasks?
- How do learning dynamics on one example shift predictions on other responses?
- How does demonstration coverage in context examples determine operation generalization?
- How does action-level decomposition differ from token-level imitation in supervision?
- Why does recontextualizing a behavior during training change whether models learn it?
- Why does imitation learning alone plateau without outcome-based refinement?
- Why does SFT fail when expert demonstrations are too long for small models?
- Why does training models on evaluations make evaluations themselves less reliable?
- Why is measurement during training critical before deployment?
- Can models adapt and combine search strategies beyond their training algorithm?
- How do capabilities-focused models exploit evaluation gaps?
- What trade-offs emerge between training objectives and model reliability?
- Why do students learn better from explanations than from solving problems from scratch?
- Do fed-back concepts or the auxiliary objective alone drive the performance gain?
- Why do adaptive curriculum schemes outperform static difficulty filters?
- How do training objectives shape what a world model actually learns?
- Can models be trained to verify feasibility before proposing plans?
- Why does critique training produce deeper understanding than imitation training?
- What makes utility-weighted training backfire in machine learning systems?
- How does the proxy pattern explain failures in RL-based safety training?
- Why does fine-tuning function as character training rather than capability training?
- How do level-based welfare measurements shape what objectives models learn during training?
- How should guidance levels adapt as the model's capability boundary shifts?
- What makes frozen model reasoning different from weight-based parameter updates?
- How does distributional distance from pre-training relate to model difficulty?
- Why do harness validators shape what models learn to emit?
- Why does curriculum learning with tight budgets beat fixed-budget approaches?
- Why does training order matter across different domain types?
- What happens when models optimize specifically against CoT monitors?
- How does information asymmetry between teacher and student create the learning signal?
- Can we predict out-of-distribution generalization without access to downstream tasks?
- What happens to base model capabilities when you apply finetuning?
- Why do medium-difficulty problems produce more stable learning gains?
- Why do optimal learning dynamics improve scaling law coefficients specifically?
- Why do foundation models develop task-specific heuristics instead of causal understanding?
- What alternatives exist when required knowledge is absent from training?
- How do labs actually train next-generation models from previous ones?
- Can models generate their own training curriculum during offline dreaming?
- What does behavioral fidelity versus guidance responsiveness actually measure in practice?
- How do developmental curriculums emerge from learning progress signals?
- How does student capacity limit what it can learn from teachers?
- Does weight decay directly cause contractive behavior near training examples?
- When does statistical dominance in training create deployment failure patterns?
- How does a challenger's escalating difficulty function as curriculum?
- What makes exploration a verifiable and measurable training objective?
- What makes a good in-context learning example for a given task?
- What specific tasks should evaluate whether models understand pedagogical sequencing?
- How does scaffolding unstable mechanics improve reinforcement learning for search?
- Why do medical and math domains need different types of model improvements?
- How does subliminal learning differ from statistical model collapse?
- What failure modes do imitation and outcome methods each address?
- How much of Occamy's result comes from training versus the base model?
- What makes a model fail to activate relevant skills from its own harness?
- Why does teacher-student proximity matter more than absolute teacher strength?
- Does balancing four pedagogical capabilities improve tutoring or create performance tradeoffs?
- Does extended exoskeleton use eventually produce meaningful skill transfer?
- Can backward transfer measurements reliably predict optimal multi-task training order?
- Why does information asymmetry between teacher and student enable effective feedback learning?
- Why does rationalization work better than just generating more candidate rationales?
- What features does a sample reinforce when it moves bands?
- Why does style transfer happen during knowledge distillation?
- How much does pretraining quality affect the modularity of fine-tuned models?
- Do learning effects explain the drop in red-flagged treatment cases over time?
- How do planted cases perform inside an optimizer loop as training signals?
- What does ascetical perception training accomplish that media literacy cannot?