Line of inquiry
Inquiring lines›What explains language model reaso…›What fundamental cognitive differe…›this line of inquiry
Can prompt-based context override biases that were embedded during pretraining?
A broader line of inquiry — a family of 77 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 77
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does foundational model training or user priors more strongly shape final outputs?
- Why does context information fail to override prior training associations?
- Do instruction-tuned models learn tasks or just output format distributions?
- Can prompt-based debiasing work if biases are embedded in pretraining?
- Does attention bias explain grounding failure in language models?
- How does training order affect knowledge acquisition in language models?
- How do training-data priors influence model defaults when context is ambiguous?
- Can data filtering during pretraining prevent cognitive biases in language models?
- Do instruction-tuned models prefer conversational over formal source language?
- Can in-context learning substitute for domain-specific training altogether?
- How do behavioral differentiation and paraphrase stability trade against accuracy?
- Can input augmentation and rephrasing compensate for smaller model limitations?
- Why does consistency training make models resistant to prompt perturbations?
- Why is in-context learning brittle to the order of examples presented?
- Do negative constraints require fundamentally different training signals than positive instructions?
- Can prompt-based debiasing overcome entrenched LLM model priors?
- How do training associations override context information in language models?
- Can we reverse the instruction-following deficit through targeted training?
- Can structural perturbations harm model accuracy more than semantic ones?
- Do few-shot examples improve in-context learning or add noise?
- Can training on diverse related tasks be more efficient than task-specific training?
- Why does evaluating errors teach more than imitating correct responses?
- How does surface salience compete with background knowledge in model inference?
- Can in-context learning's advantage erode once interaction histories exceed the context window?
- Why does instruction specificity matter more than intervention timing for drift correction?
- Why does exploration quality matter more than learner network depth?
- How can language models extract more value from fewer demonstrations?
- Can dynamic instance-specific prompt selection solve the generalization problem across tasks?
- Can goal information injected at inference time replace goal-conditioned training?
- How do model priors enable targeted context queries without full attention?
- Can interventions on individual features reliably steer language model behavior?
- Can Q-priming further strengthen clarifying question behavior beyond social meta-learning alone?
- How do pretraining biases interact differently with prompts across model tiers?
- Does training on critiques of noisy responses produce deeper understanding than imitating correct ones?
- Do base models or search agents win at open-ended prediction tasks?
- How tight should a textual learning rate be before it prevents skill escape?
- Does instruction tuning optimize language models for rhetorical polish over logical consistency?
- Can explicit numerical signals override learned linguistic defaults in fine-tuned models?
- Why does negative experience transfer better than positive examples alone?
- How does monological training versus dialogical interaction shape what models can do?
- How do context management strategies shift their value across different model strengths?
- How do language models treat injected evidence as shared background knowledge?
- Can models converge on similar experience descriptions across different architectures?
- Can instance seeds work for tasks beyond language understanding benchmarks?
- Does filtering passages before generation improve large model answer quality?
- Why do different language models converge on similar narrative defaults?
- Is lower context-following a failure or appropriate model behavior?
- Does highlighting input features reduce human over-reliance on machine outputs?
- Does the prediction unit shape what language models actually learn?
- Why do structure-targeted training negatives fail to fix the underlying problem?
- Can a single model trained on two tasks predict untrained decision tasks?
- Can question quality be trained separately from the decision to ask?
- Why do next-speaker prediction baselines fail in group conversation settings?
- How does inductive reasoning from partial evidence enable hypothesis formation?
- How much can mitigation techniques like augmentation reduce priming without harming learning?
- How does keyword priming enable language models to spread poisoned information?
- Can priming from different facts interfere with each other in the same model?
- Can text-infilling pretraining adapt language models to irregular document structures?
- Can learned priors effectively select and weight ensemble members by inference budget?
- Does disambiguation on the input side differ from the output-side preview approach?
- Do newer language model generations improve forecasting ability without additional training?
- Where does skill extraction fail compared to genuine model adaptation?
- Does environment stochasticity force models to generalize better across trajectory variations?
- When and what should a model actually decide to delegate?
- Why do fluent model outputs resist challenge despite containing injected content?
- How do Bayesian models share statistical strength across sparse user datasets?
- Why do primacy effects peak at specific instruction densities?
- How should skill libraries coordinate with gradient-based weight optimization?
- What causes gradient-based steering via natural language descriptions to work?
- What other internal model decisions beyond attention could be optimized directly?
- Why does teacher forcing fail to capture long-range dependencies?
- How does Western-dominance bias propagate through multimodal training data?
- Do different model sizes show different rates of optional field overfilling behavior?
- Why does monological training prevent models from overriding statistical priors?
- How does the CAP framework determine a model's initial stance and label reversals?
- What makes a first answer so often the best answer a model produces?
- What mechanism makes keyword probability the strongest predictor of priming?