Do language models really understand meaning or just surface frequency?
Explores whether LLMs comprehend semantic meaning independently of textual frequency, or whether high-frequency paraphrases systematically outperform rare ones even when meaning is identical across math, translation, and reasoning tasks.
Adam's Law (TFL) generalizes a previously local finding into a global property of LLM computation. The earlier NLI work showed predicates in entailment hypotheses skew higher-frequency than premises, and that fine-tuning amplifies rather than dilutes this bias — see Does fine-tuning on NLI teach inference or amplify shortcuts?. Adam's Law extends this across four task families: math reasoning, machine translation across hundreds of language pairs, commonsense reasoning, and agentic tool calling. The constant: when meaning is held fixed and only surface form varies, the higher-frequency paraphrase outperforms the lower-frequency one.
The mechanism is straightforward but uncomfortable. Higher-frequency text occurred more often during pre-training, so it sits in a denser, better-modeled region of the distribution. The model's "comprehension" is therefore not meaning-recognition first and surface-decoding second — it is statistical-mass recognition first, with meaning emerging downstream of that recognition. This converges with Can models pass tests while missing the actual grammar?: correct outputs do not certify that meaning is what the model is tracking.
The pattern matters because paraphrase invariance is a load-bearing assumption almost everywhere LLMs are deployed. We assume the same prompt, said two ways, will yield the same answer. Adam's Law says no: it will yield the frequency-weighted answer, and the surface form is a covariate of accuracy, not a transparent vehicle for the request. This also shadows the output side. Do different AI models actually produce diverse outputs? documents convergence in what models say; Adam's Law documents the same convergence in how models comprehend what is said to them. Both endpoints of the prompt-response loop pull toward the corpus mean. Frequency is not noise around meaning. Frequency is a substantial fraction of what comprehension means inside a transformer.
Inquiring lines that read this note 91
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.
Why do multiple language models independently generate similar outputs? Can we reliably distinguish AI-generated from human-written text?- Can AI detect sense-of-nonsense the way human readers do?
- Can adversarial paraphrasing defeat feature-based detection of LLM text?
- How does propositional density differ between dense and sparse generated text?
- Does the semantic weight of AI-written content matter more than sentence count?
- How much does model-specific phrasing contribute to detectable style versus semantic content differences?
- Why does training data saliency distort how models judge meaning?
- Can language models acquire meaning from distributional patterns alone without joint attention?
- Does generalization frequency explain why models favor upward semantic movement?
- What distinguishes surface cues from structural meaning in language understanding?
- Can benchmark performance distinguish surface from structural linguistic knowledge?
- Do LLMs learn linguistic generalizations or just surface-level frequency patterns?
- How do language models transmit traits through semantically unrelated data?
- What distinguishes real understanding from superficial pattern matching?
- Can autoformalisation from natural language preserve semantic accuracy?
- What makes ambiguity recognition fundamentally important for poetry analysis?
- What percentage of natural language relies on plausible deniability through ambiguous phrasing?
- How does semantic ambiguity differ from structural ambiguity in language?
- What other semantic relations benefit from explicit surface markers in text?
- What semantic classifier design avoids lexical variation without genuine conceptual distinctness?
- Can frame semantics explain why context matters more than word similarity?
- Can adding more words to a passage actually interfere with meaning?
- Why do different readers extract different meanings from identical text?
- Why does joint attention matter for acquiring linguistic meaning?
- How do static embeddings and contextualized representations divide semantic labor?
- Does language convey meaning purely through relational structure without external grounding?
- Can readers detect meaning through resonance patterns alone without knowing authorial intent?
- Where does the meaning actually originate in reader-detected resonance across language?
- Does functional grounding through discourse patterns count as genuine semantic meaning?
- Can language models ground clarifications without vision and kinesthetic modalities?
- How does semantic grounding differ between human minds and language models?
- Can language models develop world models that ground meaning in causal reality?
- Why does combining natural language with numerical scores improve prediction accuracy?
- Can meaning-level metrics like Semantic Entropy avoid length bias?
- How does syntactic encoding relate to semantic feature representation?
- Why do language models fail when semantic content is stripped away?
- How do rare linguistic registers differ from conceptually complex examples?
- Can LLMs improve at metaphor if they handle decoupled semantics better?
- Do LLMs struggle more with semantic accuracy than syntactic correctness across domains?
- Why do LLMs fail at semantic generalization despite grammatical accuracy?
- How does the distance between natural language and formal notation affect translation accuracy?
- Why do LLMs choose surface-order quantifier scope over contextually correct readings?
- Can LLM semantic representations exist without causally influencing their generation output?
- How much semantic meaning survives when LLMs paraphrase poetry and literary text?
- How do LLMs compress literary language without losing essential nuance?
- Why does probability of text completion not equal knowledge value?
- Can language models translate theorems faithfully without semantic loss?
- Is interpretive multiplicity a bug in language or a feature?
- Do LLMs compute scalar implicature differently across conversational contexts?
- How does implicit meaning processing limit LLM pragmatic reasoning?
- Is paraphrase invariance a reliable assumption when deploying language models in production?
- Do metaphors work by decoupling meaning from linguistic associations?
- Can presupposition projection strength vary by context in embeddings?
- How does bidirectional entailment distinguish semantic equivalence from token similarity?
- How should meaning spaces be systematically modeled across different applications?
- Can encoder models match human conceptual structure better than larger language models?
- Why do NLP models fail at recognizing multiple valid interpretations?
- How does modeling capability relate to lossless compression in language models?
- What is the comprehension-generation asymmetry in language models?
- Why do embeddings measure semantic association instead of task relevance?
- What makes vector embeddings fail on single-hop semantic relevance queries?
- Why do semantic similarity and task relevance diverge in vector search results?
- Can vector embeddings measure task relevance instead of semantic similarity?
- How well does semantic similarity preserve survey response nuance?
- What paraphrase and conceptual matching tasks favor dense over exact-match retrieval?
- Can embedding similarity reliably map language model outputs to survey response scales?
- Can semantic search find paraphrased and renamed tasks without human review?
- Why does homework adherence remain low despite advances in language model capability?
- Why do speech benchmarks still measure transcription instead of comprehension?
- Can LLMs identify implicit metaphoric mappings that require pragmatic inference?
- Do LLMs track surface wording more than semantic meaning in moral judgment?
- What distinguishes conceptual understanding from statistical pattern matching in models?
- How do mechanistic features compare to natural language for interpretability?
- Does model confidence actually explain why paraphrases produce different outputs?
- How do behavioral differentiation and paraphrase stability trade against accuracy?
- Why does cross-text analogical reasoning fail when semantics decouple from symbols?
- Why do readability and style metrics plateau while reasoning improves with scale?
- Why does training data not function as a searchable corpus?
- How does training distribution shape what language models understand best?
Related concepts in this collection 3
This note in its neighbourhood — explore the map, then jump to a related concept in the list below.
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
-
Does fine-tuning on NLI teach inference or amplify shortcuts?
When LLMs are fine-tuned on natural language inference datasets, do they learn genuine reasoning abilities or become better at exploiting statistical patterns in the training data? Understanding this distinction matters for assessing model capabilities.
local finding that Adam's Law generalizes
-
Can models pass tests while missing the actual grammar?
Do language models succeed on grammatical benchmarks by learning surface patterns rather than structural rules? This matters because correct outputs may hide reliance on shallow heuristics that fail on novel structures.
mechanism: surface, not semantics
-
Do different AI models actually produce diverse outputs?
Explores whether using multiple different language models together creates genuine diversity or whether shared training and alignment cause them to converge on similar answers despite independence.
output-side counterpart of the same dynamic
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- Adam's Law: Textual Frequency Law on Large Language Models
- Beyond Accuracy: Evaluating the Reasoning Behavior of Large Language Models -- A Survey
- Word Meanings in Transformer Language Models
- Simple Linguistic Inferences of Large Language Models (LLMs): Blind Spots and Blinds
- Six misconceptions about large language models: A minimal model and diagnostic taxonomy
- Bigger is not always better: The importance of human-scale language modeling for psycholinguistics
- Farther the Shift, Sparser the Representation: Analyzing OOD Mechanisms in LLMs
- Language models show human-like content effects on reasoning tasks
Original note title
high-frequency phrasing wins — LLMs systematically prefer textually frequent paraphrases over rare ones with the same meaning