SYNTHESIS NOTE
TopicsNatural Language Inferencethis note

Do LLMs predict entailment based on what they memorized?

Explores whether language models make entailment decisions by recognizing memorized facts about the hypothesis rather than reasoning through the logical relationship between premise and hypothesis.

Synthesis note · 2026-02-21 · sourced from Natural Language Inference
What kind of thing is an LLM really? How do you navigate synthesis across fragmented research topics?

McKenna et al. (2023) named a specific, reproducible bias in LLM entailment behavior: the attestation bias. When an LLM is asked whether premise P entails hypothesis H, its prediction is bound to the hypothesis's out-of-context truthfulness — whether H is attested in training data — rather than the conditional truth of H given P.

The mechanism is clear: if a model's training data confirms H as true (independently of any premise), the model is likely to predict entailment regardless of what P says. Conversely, if H is not attested, the model is less likely to predict entailment even when it would be correct. Entities serve as "indices" to memorized propositions — the presence of a known entity activates stored associations that override the in-context reasoning task.

The authors demonstrate this with a "random premise" experiment: replace the original premise with a random unrelated premise while keeping H constant. An ideal inference model should detect that entailment is no longer supported and predict "no entailment." LLMs instead maintain elevated entailment predictions when H is attested — demonstrating that they are responding to stored propositions about H, not to the P→H relationship.

This connects to two complementary failure modes already in the vault. Do language models actually use their encoded knowledge? shows that encoded knowledge doesn't reliably affect generation. Attestation bias is the inverse problem: memorized statements do influence generation, but in the wrong direction — they substitute for rather than support proper inference. Both failures arise from the same root: LLM generation is not governed by a clean separation between retrieved knowledge and in-context reasoning.

The practical implication: NLI benchmark performance measures a combination of reasoning and memorization that cannot be cleanly disentangled without carefully designed bias-adversarial test sets.

Inquiring lines that read this note 54

This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.

Can AI-generated outputs constitute genuine knowledge or valid claims? How do language models establish social grounding in human dialogue? Can prompting inject entirely new knowledge into language models? How can models identify insufficient information and respond appropriately without guessing? Do language models learn genuine linguistic structure or just surface patterns? How should retrieval systems optimize for multi-step reasoning during inference? How do training priors constrain what context information can override? Do language models perform faithful symbolic reasoning independent of semantic grounding? How faithfully do LLMs reflect their actual reasoning in outputs and explanations? How do language models inherit human biases from training data? How do training data properties shape reasoning capability development? Can prompting strategies overcome LLM biases without model fine-tuning? Why do language models struggle with implicit discourse relations? Why do continual learning scenarios trigger catastrophic forgetting and interference? How does memorization interact with learning and generalization? Why does supervised fine-tuning improve accuracy while degrading reasoning quality? Do language models understand semantics or rely on pattern matching? Why do language models reinforce false assumptions instead of correcting them? How do LLMs distinguish causal reasoning from temporal and semantic associations? Why do reasoning models fail at systematic problem-solving and search? Do base models contain latent reasoning that training can unlock? Do accurate-looking LLM outputs hide structural failures in learning and reasoning? How do evaluation biases undermine LLM quality assessment systems? Why does finetuning cause catastrophic forgetting of model capabilities? How do neural networks separate factual knowledge from reasoning abilities? Does model scaling alone produce compositional generalization without symbolic mechanisms?

Related concepts in this collection 3

This note in its neighbourhood — explore the map, then jump to a related concept in the list below.

Concept map
15 direct connections · 176 in 2-hop network ·dense cluster Open in graph ↗

Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph

your link semantically near linked from elsewhere

Related papers in this collection 8

Papers most semantically related to this note, ranked by cosine similarity in the embedding space.

Original note title

llm entailment predictions are bound to hypothesis attestation rather than premise-hypothesis inference