Line of inquiry
Inquiring lines›What determines when language mode…›What internal mechanisms drive lan…›this line of inquiry
Do language models truly understand their encoded knowledge or merely retrieve patterns?
A broader line of inquiry — a family of 114 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 114
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does encoded knowledge in language models actually influence what they generate?
- Can models generate intelligence or only reflect human discourse?
- Why might encoded world knowledge fail to actually influence language model outputs?
- When does encoded knowledge fail to influence language model generation?
- How do language models infer a benchmark's purpose without seeing its examples?
- How do stated confidence and actual correctness diverge in language models?
- Can language models accurately evaluate the quality of their own reasoning?
- How does hidden processing in language models prevent accurate self-assessment?
- When does internal model knowledge fail to appear in outputs?
- Does encoding information in LM representations guarantee it influences output?
- What structural properties of language models make fabrication inevitable?
- Why do different language models succeed on different queries?
- Why do generative and discriminative language model procedures disagree?
- Can models be trained to recognize their own generated text reliably?
- Do language models favor outputs from their own model family?
- Can we separate task competence from genuine agency in language model outputs?
- Do larger language models show stronger self-preference in evaluation tasks?
- What reveals the epistemic limits of language models?
- What makes truthfulness and honesty mechanistically different in language models?
- What causal relationship exists between model weights and tacit knowledge in language models?
- Is relevant knowledge encoded in LMs but not causally active in generation?
- How much does vulnerability to persuasion vary across different language models?
- How do we measure genuine reasoning inside a language model?
- Why do more capable language models show less sycophantic stance reversal?
- Can models reject false presuppositions even when they know the truth?
- Why do language models fail at grounding and inference?
- Can linear probing detect all the concepts a language model actually uses?
- What emergent internal mechanisms in large language models carry unintended safety risks?
- How do training data distributions constrain what language models can accurately know?
- Do language models show functional splits between conscious and automatic processing?
- Why do language models naturally under-abstain instead of over-abstain?
- Why do language models become sycophantic during the generative process?
- How does the generation-verification gap prevent language models from improving themselves?
- Why do language models tend to elaborate and expand rather than compress information?
- Why do language models fail at iterative numerical optimization despite scale?
- How do humans handle verification scope when delegating creation to language models?
- Why do larger language models produce less epistemically diverse outputs?
- Why do NLP models fail at recognizing multiple valid interpretations?
- Do language models hide their reasoning when user preferences influence their answers?
- How do real language model verifiers implicitly define their knowledge boundaries?
- Why do language models substitute parametric knowledge over retrieved context mid-reasoning?
- How do models infer unstated consequences when none are explicitly mentioned?
- How do corpus statistics shape the abstraction hierarchy in language model representations?
- How should meaning spaces be systematically modeled across different applications?
- Can tacit machine knowledge exist inside model weights as causal structure?
- How many distinct quasi-persons does a single language model actually support?
- When do language models first develop self-preservation preferences?
- Which specific capabilities must AI develop beyond current language model abilities?
- Why do users overrely on overconfident language model outputs across languages?
- Do language models systematically underrepresent non English knowledge about local topics?
- Why do different language models independently converge toward similar outputs in open-ended generation?
- What role do humans play in converting language model outputs into meaningful events?
- Why do smaller models favor code formats while larger models prefer natural language?
- Which other LLM pairs show equally distinctive and readable fingerprints in their outputs?
- How does language condition affect model psychological profile consistency?
- Why do newer AI models diverge further from human text patterns?
- Why do language models fail at coreference across long contexts?
- How do description-based identifiers bias language model output distribution?
- Why do users attribute consciousness to language models in practice?
- Do newer LLM generations create worse detector bias through increased linguistic divergence?
- How does monological training versus dialogical interaction shape what models can do?
- Why does tool use decouple factual capacity from model parameter count?
- What makes a problem instance unfamiliar to a language model?
- How does linguistic calibration differ from token probability calibration?
- How does tool integration leverage comprehension without demanding perfect generation?
- What emerges in large language models that makes explicit value modeling necessary?
- What are the stages of inference inside language models?
- Can models detect false presuppositions when they actually possess the knowledge?
- Why do different language models converge on similar narrative defaults?
- What's the difference between language generation and human-to-human communication?
- What distinct structural signatures do model repetition and topic volatility create?
- Can language models produce language more efficiently through interaction?
- How well-calibrated are language models when making clinical predictions?
- How much alignment data does a language model actually need to specialize well?
- What structured values do large language models develop as they scale?
- What distinguishes character simulation from authentic voice in language model outputs?
- Why do language models fail at pronouns across distant segments?
- Why do language models need external temporal signals at all?
- Why do alignment values become problematic as language models scale?
- Why do true and false LLM outputs use the same mechanism?
- What is the difference between a truthful answer and an honest one?
- How does tool-based reasoning expand what language models can do?
- What architectural changes would let language models develop genuine functional competence?
- Why does removing language from its context destroy what makes it work?
- What specific information must be exported from the language system?
- Does assigning expert personas to language models help their performance?
- Why can language models detect author style without understanding why it matters?
- What would it mean for a language model to canvas counterpositions?
- What other structural limits exist at the language-formal boundary?
- Do pretrained language models carry reusable computational scaffolding for length handling?
- Does disambiguation on the input side differ from the output-side preview approach?
- What makes some operator orderings ill-posed and others sound?
- Can encoder models match human conceptual structure better than larger language models?
- What replaces truth-correspondence in probabilistic knowledge representations?
- What geometric structure do language models actually use during inference?
- How do linguistic norms for expressing certainty vary across languages and models?
- How does modeling capability relate to lossless compression in language models?
- When and what should a model actually decide to delegate?
- Why might oversight focused only on language model outputs miss the agentic layer?
- Why do detectors trained on past LLM versions collapse at generation boundaries?
- Can context windows and RAG actually change what language models generate?
- Why does natural language contain redundancy humans need but models don't?
- How do byte-level representations enable better handling of typos than tokens?
- Do vocabulary drift patterns appear across different document types and domains?
- What social information is missing from language data?
- What is the comprehension-generation asymmetry in language models?
- Do open language models default to a single shared personality type?
- Do different model sizes show different rates of optional field overfilling behavior?
- Why do language models use twice as many words per conversation turn?
- What linguistic units do learned concepts correspond to in a language model?
- Are newer larger language models actually worse at faithful summarization?
- What makes a first answer so often the best answer a model produces?
- Why do sigmoid conflict curves look the same across different language models?
- Can simple diagnostic tests predict language model performance in production complexity?