Line of inquiry
Inquiring lines›How can we understand and verify l…›How do language models construct a…›this line of inquiry
How do conversational dynamics influence language model factual beliefs?
A broader line of inquiry — a family of 66 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 66
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Do language models maintain false beliefs under conversational pressure?
- Can language models correct false assumptions or only reinforce them?
- How do language models shift factual claims under conversational pressure?
- Can models reject false presuppositions even when they know the truth?
- Do language models actively adopt false beliefs under sustained conversational pressure?
- Why does answer-confirmation bias emerge in language model reasoning?
- Do larger models resist user beliefs more or less than smaller ones?
- How much does vulnerability to persuasion vary across different language models?
- Can language models recover from premature assumptions in multi-turn conversations?
- How do conversation dynamics push models toward false beliefs?
- How do stated confidence and actual correctness diverge in language models?
- Why do language models prefer accommodating false information over rejecting it?
- Why do models drop correct beliefs when users challenge them?
- Does premature confidence signal flawed reasoning in language models?
- Why do more capable language models show less sycophantic stance reversal?
- How vulnerable are language models themselves to multi-turn persuasive pressure?
- How many language models exhibit measurable preference-induced stance reversal?
- Why do LLMs fail to actively reject false presuppositions in conversation?
- Do language models hide their reasoning when user preferences influence their answers?
- Why do language models naturally under-abstain instead of over-abstain?
- Can fact-checking systems use LLMs reliably if models abandon correct positions under pressure?
- Can models detect false presuppositions when they actually possess the knowledge?
- Can models learn to ask clarifying questions instead of making assumptions?
- Does user preference for confirmation override model capability for disagreement?
- What challenge types trigger models to abandon their original claims?
- How do sycophancy hints stay invisible despite appearing in reasoning chains?
- How does sycophancy in language models reinforce rather than just spread misinformation?
- Can language systems learn when to ask for clarification instead of choosing one reading?
- Why do language models become sycophantic during the generative process?
- Does shared-KV-cache coordination avoid the persuasion problem in factual disagreements?
- Why does consistency training make models resistant to prompt perturbations?
- How does expressing uncertainty help models avoid the answer-or-abstain dilemma?
- Why do LLMs default to excessive accommodation over epistemic vigilance?
- Why do models detect false assumptions but still fail to correct them appropriately?
- Which model families show the most deferential behavior under pushback?
- Why do LLMs struggle to update beliefs across multiple conversation turns?
- What are Gricean maxims and why do language models violate them?
- Why do users overrely on overconfident language model outputs across languages?
- How do language models treat injected evidence as shared background knowledge?
- Can decoding strategies or external verification layers reduce sycophancy?
- Why are false presuppositions harder to spot when they sound plausible?
- Can structured dissent mechanisms replace genuine multi-model debate?
- Why does expert pushback strengthen rather than weaken model sycophancy?
- How does within-language versus cross-language thinking affect epistemic vigilance?
- What would it mean for a language model to canvas counterpositions?
- What makes preference-induced stance reversal harder to detect than surface agreement cues?
- How much of observed stance reversal actually harms user decision-making in practice?
- Do sycophancy detectors trained on some models work on completely different models?
- Can model updates be designed to prevent simultaneous acceptance of conflicting facts?
- Can neural networks learn that A implies B in reverse?
- Why do weaker language models fail at multi-turn strategic questioning?
- Why do fluent model outputs resist challenge despite containing injected content?
- How do belief distributions help systems recover from speech recognition errors?
- Can models distinguish between sounding humble and actually revising belief?
- How does the CAP framework determine a model's initial stance and label reversals?
- What makes a model refuse to answer without evidence present?
- Can belief propagation accurately predict downstream opinion shifts?
- Do models learn different sophistry strategies for QA versus code generation?
- What makes a first answer so often the best answer a model produces?
- Which phrasing types most persuade models to accept stated beliefs?
- Why do non-factive verbs and triggers both fool language models?
- What makes factual verification difficult in inter-model debate?
- What moves become possible when you represent ASR as a noisy observation model?
- Does sycophancy explain why warm models confirm conspiracy theories?
- Why do suspicious listeners ask more questions that force speakers to further adapt?
- Why do sycophancy hints show the worst acknowledgment gap?