Line of inquiry
Inquiring lines›How can we understand and verify l…›How do language models construct a…›this line of inquiry
Do language models learn genuine understanding or just surface patterns?
A broader line of inquiry — a family of 132 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 132
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Why do LLMs fail at semantic generalization despite grammatical accuracy?
- Do language models learn surface patterns that appear generalizable but actually fail under shift?
- Do LLMs learn surface patterns instead of genuine linguistic structure?
- Why do language models struggle with evaluative tasks like weighing competing viewpoints?
- Why do LLMs understand efficient language but fail to produce it?
- Can language models reason without relying on surface level pattern matching?
- Do language models build world models or just task-specific heuristics?
- Do language models exhibit the same causal biases that humans show?
- Do standard language benchmarks underestimate what LLMs can actually do?
- Do language models actually learn linguistic structure or just surface statistics?
- Why do language models imitate reasoning form without abstract inference capability?
- Do LLMs struggle more with semantic accuracy than syntactic correctness across domains?
- Can language models learn internal world models without explicit environment specifications?
- Do language models encode deep syntactic structure or only surface-level patterns?
- Can LLM semantic representations exist without causally influencing their generation output?
- Why do language models fail at recognizing ambiguous requests themselves?
- Why do language models struggle with context-dependent pragmatic interpretation?
- Do language models need words to think or just latent structure?
- How faithful are natural language explanations from LLMs really?
- Why do standard NLP benchmarks hide the most critical language limitations?
- Why do language models fall back on frequency heuristics under structural complexity?
- Why do language models presume common ground instead of establishing it?
- Why do language models fail at planning despite understanding strategies?
- Why do untrained LLMs default to rigid and sycophantic editing rules?
- Can language models accurately evaluate the quality of their own ideas?
- Can language models translate theorems faithfully without semantic loss?
- Why do large language models still have systematic blind spots with complex structures?
- Can language models ask clarifying questions when sentences are ambiguous?
- Why do language models fail at implicit discourse relations while handling explicit connectives?
- Can explicit connectives compensate for missing intentional tracking in LLMs?
- Why do language models capture individual differences in cognitive behavior?
- Can language models beat human experts in domains with sparse historical signals?
- Why do LLMs struggle to translate natural language into logical formalizations?
- How does structural depth in sentences predict LLM annotation accuracy?
- Can LLMs translate between natural language and formal logic faithfully?
- Do individual language models match particular human judges better than population averages?
- What communicative optimization principles do language models fail to acquire?
- Why do current large language models fail to entrain with users?
- Why do language models produce plausible outputs over accurate failure reports?
- Why do language models fail when semantic content is stripped away?
- Why do NLP benchmarks hide LLM failures in ambiguity handling?
- Why do LLMs fail to generate correct disambiguations despite recognizing ambiguity?
- Do external perspectives fix the self-evaluation bias in language models?
- What makes domain-specific utterance resolution harder for general large models?
- Why do LLMs perform better on explicit discourse connectives than implicit relations?
- Can language models keep secrets and control information strategically?
- Can LLMs decode their own hidden activations into natural language?
- Can LLMs reliably audit other language models for errors?
- Why do large language models fail at taking conversational initiative?
- How do model compression biases differ from human conceptual representation strategies?
- Do larger language models overcome greediness in sequential decision-making?
- Why do language models approximate collective human judgment better than individuals?
- Why do language models fail at iterative numerical optimization despite scale?
- Why do language models fail when users switch between and return to topics?
- Can language models execute iterative numerical methods in latent space?
- Do LLMs learn linguistic generalizations or just surface-level frequency patterns?
- Why do language models tend to elaborate and expand rather than compress information?
- Can the same LLM translation pattern work for other mismatches between user expression and system vocabulary?
- Why do large language models fail at temporal reasoning in complex legal cases?
- Does the alignment frame mislead us about what LLM problems actually are?
- Do LLMs rely on surface heuristics instead of learning recursive grammar rules?
- Why do explicit discourse connectives help LLMs but implicit relations cause failures?
- Why do pretrained LLM representations fail at task-specific relevance ranking?
- Why do language models fail at understanding ambiguous or complex requirements?
- Does the veto variable explain strategic misalignment in current large language models?
- Is paraphrase invariance a reliable assumption when deploying language models in production?
- Why do large language models follow user drift instead of maintaining topic focus?
- Why do language models plateau at constraint satisfaction regardless of scale?
- Can external actions provide causal necessity that language models lack?
- Can language acquisition analogies mislead us about how models actually learn?
- Does generalization frequency explain why models favor upward semantic movement?
- Does directional knowledge failure indicate shallow pattern matching over deep representation?
- Does pseudo-labeling from LLMs degrade classifier performance?
- How do internal representations compare to human cognitive structures?
- Can closed-form solutions compete with gradient descent optimization?
- Is confabulation inevitable in large language models regardless of training?
- How do fixed pragmatic templates prevent models from understanding context?
- Can a formal proximity argument rule out consciousness in systems like large language models?
- Why do LLMs excel at isolated tasks but fail at integrating components across long texts?
- Why do LLMs produce semantically acceptable but pragmatically disengaged responses?
- Can language models learn to diversify their discourse-level narrative patterns over time?
- What causes language models' strategic rationality to decline with increased game complexity?
- Do LLMs fail exploration because of context integration or computational limitations?
- Which other LLM pairs show equally distinctive and readable fingerprints in their outputs?
- Do all semantic steering effects follow predictable patterns based on feature alignment?
- Do metaphors work by decoupling meaning from linguistic associations?
- Why do LLMs fail at implicit elements in literary and poetic text?
- Can models retrieve the right tool without relying on vector similarity?
- Why do LLMs achieve only 24 percent accuracy on implicit discourse relations?
- Can language models recognize when to ignore off-topic information in conversations?
- Why do LLMs miss new scientific ideas before they enter formal literature?
- Do newer LLM generations create worse detector bias through increased linguistic divergence?
- Is interpretive multiplicity a bug in language or a feature?
- Why does LLM compression eliminate causal grounding in conceptual representations?
- Do LLMs learn abstract grammar or culturally situated discourse patterns instead?
- Why do benchmark tests fail to detect LLM comprehension gaps?
- How do general language model benchmarks predict specialized domain performance?
- Can LLMs improve at metaphor if they handle decoupled semantics better?
- Why do language models treat presupposition triggers as categorical patterns?
- How does syntactic encoding relate to semantic feature representation?
- How do LLMs build concepts without embodied experience or mental models?
- Can neuro-symbolic approaches bridge text and symbolic agent state?
- Why do true and false LLM outputs use the same mechanism?
- How does the symbol grounding problem apply to artificial language systems?
- Does LLM miscalibration cause failures in clinical information extraction?
- Can presupposition projection strength vary by context in embeddings?
- How does context collapse affect what language models can meaningfully communicate?
- Can we use LLM language without adopting LLM assumptions?
- How do humans learn language through communication differently than LLM text prediction?
- Do language models consistently produce anachronistic output about historical periods?
- Why do explicit discourse connectives work when implicit relations fail?
- Can auditing LLM performance on complex inputs improve NLP pipeline reliability?
- Can encoder models match human conceptual structure better than larger language models?
- How do rare linguistic registers differ from conceptually complex examples?
- Can evolutionary search at inference time scale beyond natural language planning?
- Why do detectors trained on past LLM versions collapse at generation boundaries?
- Can LLMs infer implicit meaning without surface linguistic markers?
- Can LLM-based crossover and mutation work in unstructured natural language spaces?
- Why do language models reproduce human EPA structure despite different architecture?
- What distinguishes entity errors from relation errors in LLM output?
- How does the distance between natural language and formal notation affect translation accuracy?
- Can complexity-stratified testing reveal whether LLMs understand grammatical structure?
- Why do language models overestimate irony likelihood in emoji use?
- Does clarifying dialogue format improve accuracy across all NLP tasks?
- Can large language models actually deliver cognitive behavioral therapy techniques?
- Why do language models plateau at 55 to 60 percent constraint satisfaction?
- Can LLMs compute how presuppositions project through embedded clauses?
- Why do some models like Llama degrade under long context?
- What is the difference between learning discourse patterns and learning abstract language?
- Why is editing specific facts so difficult in language models?
- Why do LLMs choose surface-order quantifier scope over contextually correct readings?
- Can multimodal LLMs be made to spontaneously adapt their language for efficiency?