Matches in SemOpenAlex for { <https://semopenalex.org/work/W3166595320> ?p ?o ?g. }
- W3166595320 abstract "We propose an alternate approach to quantifying how well language models learn natural language: we ask how well they match the statistical tendencies of natural language. To answer this question, we analyze whether text generated from language models exhibits the statistical tendencies present in the human-generated text on which they were trained. We provide a framework--paired with significance tests--for evaluating the fit of language models to these trends. We find that neural language models appear to learn only a subset of the tendencies considered, but align much more closely with empirical trends than proposed theoretical distributions (when present). Further, the fit to different distributions is highly-dependent on both model architecture and generation strategy. As concrete examples, text generated under the nucleus sampling scheme adheres more closely to the type--token relationship of natural language than text produced using standard ancestral sampling; text from LSTMs reflects the natural language distributions over length, stopwords, and symbols surprisingly well." @default.
- W3166595320 created "2021-06-22" @default.
- W3166595320 creator A5029033876 @default.
- W3166595320 creator A5061951606 @default.
- W3166595320 date "2021-05-31" @default.
- W3166595320 modified "2023-09-23" @default.
- W3166595320 title "Language Model Evaluation Beyond Perplexity" @default.
- W3166595320 cites W1494946869 @default.
- W3166595320 cites W1968625547 @default.
- W3166595320 cites W1969898367 @default.
- W3166595320 cites W1974971448 @default.
- W3166595320 cites W1980299258 @default.
- W3166595320 cites W2048176942 @default.
- W3166595320 cites W2055288598 @default.
- W3166595320 cites W2064675550 @default.
- W3166595320 cites W2079656678 @default.
- W3166595320 cites W2105494575 @default.
- W3166595320 cites W2183162925 @default.
- W3166595320 cites W2478258606 @default.
- W3166595320 cites W2549835527 @default.
- W3166595320 cites W2604677735 @default.
- W3166595320 cites W2736889448 @default.
- W3166595320 cites W2798935874 @default.
- W3166595320 cites W2799087070 @default.
- W3166595320 cites W2864832950 @default.
- W3166595320 cites W2906152891 @default.
- W3166595320 cites W2950568044 @default.
- W3166595320 cites W2950858167 @default.
- W3166595320 cites W2954460650 @default.
- W3166595320 cites W2962784628 @default.
- W3166595320 cites W2962832505 @default.
- W3166595320 cites W2963341956 @default.
- W3166595320 cites W2963382396 @default.
- W3166595320 cites W2963403868 @default.
- W3166595320 cites W2963631907 @default.
- W3166595320 cites W2963651521 @default.
- W3166595320 cites W2963751529 @default.
- W3166595320 cites W2963970792 @default.
- W3166595320 cites W2964103964 @default.
- W3166595320 cites W2964222268 @default.
- W3166595320 cites W2964335542 @default.
- W3166595320 cites W2970862333 @default.
- W3166595320 cites W2996287690 @default.
- W3166595320 cites W3033347850 @default.
- W3166595320 cites W3034685497 @default.
- W3166595320 cites W3035725000 @default.
- W3166595320 cites W3118026775 @default.
- W3166595320 hasPublicationYear "2021" @default.
- W3166595320 type Work @default.
- W3166595320 sameAs 3166595320 @default.
- W3166595320 citedByCount "0" @default.
- W3166595320 crossrefType "posted-content" @default.
- W3166595320 hasAuthorship W3166595320A5029033876 @default.
- W3166595320 hasAuthorship W3166595320A5061951606 @default.
- W3166595320 hasConcept C100279451 @default.
- W3166595320 hasConcept C106131492 @default.
- W3166595320 hasConcept C137293760 @default.
- W3166595320 hasConcept C140779682 @default.
- W3166595320 hasConcept C154945302 @default.
- W3166595320 hasConcept C166957645 @default.
- W3166595320 hasConcept C195324797 @default.
- W3166595320 hasConcept C204321447 @default.
- W3166595320 hasConcept C2776187449 @default.
- W3166595320 hasConcept C2776608160 @default.
- W3166595320 hasConcept C31972630 @default.
- W3166595320 hasConcept C38652104 @default.
- W3166595320 hasConcept C41008148 @default.
- W3166595320 hasConcept C48145219 @default.
- W3166595320 hasConcept C95457728 @default.
- W3166595320 hasConceptScore W3166595320C100279451 @default.
- W3166595320 hasConceptScore W3166595320C106131492 @default.
- W3166595320 hasConceptScore W3166595320C137293760 @default.
- W3166595320 hasConceptScore W3166595320C140779682 @default.
- W3166595320 hasConceptScore W3166595320C154945302 @default.
- W3166595320 hasConceptScore W3166595320C166957645 @default.
- W3166595320 hasConceptScore W3166595320C195324797 @default.
- W3166595320 hasConceptScore W3166595320C204321447 @default.
- W3166595320 hasConceptScore W3166595320C2776187449 @default.
- W3166595320 hasConceptScore W3166595320C2776608160 @default.
- W3166595320 hasConceptScore W3166595320C31972630 @default.
- W3166595320 hasConceptScore W3166595320C38652104 @default.
- W3166595320 hasConceptScore W3166595320C41008148 @default.
- W3166595320 hasConceptScore W3166595320C48145219 @default.
- W3166595320 hasConceptScore W3166595320C95457728 @default.
- W3166595320 hasOpenAccess W3166595320 @default.
- W3166595320 hasRelatedWork W1812447209 @default.
- W3166595320 hasRelatedWork W2197913429 @default.
- W3166595320 hasRelatedWork W2251946391 @default.
- W3166595320 hasRelatedWork W2252108951 @default.
- W3166595320 hasRelatedWork W2467343197 @default.
- W3166595320 hasRelatedWork W2741642835 @default.
- W3166595320 hasRelatedWork W2883597624 @default.
- W3166595320 hasRelatedWork W2952718809 @default.
- W3166595320 hasRelatedWork W2995273010 @default.
- W3166595320 hasRelatedWork W2997012196 @default.
- W3166595320 hasRelatedWork W3028787912 @default.
- W3166595320 hasRelatedWork W3046003314 @default.
- W3166595320 hasRelatedWork W3092675073 @default.
- W3166595320 hasRelatedWork W3102226577 @default.
- W3166595320 hasRelatedWork W3122021937 @default.