Matches in SemOpenAlex for { <https://semopenalex.org/work/W2952208026> ?p ?o ?g. }
Showing items 1 to 96 of
96
with 100 items per page.
- W2952208026 abstract "Recent work has shown that LSTMs trained on a generic language modeling objective capture syntax-sensitive generalizations such as long-distance number agreement. We have however no mechanistic understanding of how they accomplish this remarkable feat. Some have conjectured it depends on heuristics that do not truly take hierarchical structure into account. We present here a detailed study of the inner mechanics of number tracking in LSTMs at the single neuron level. We discover that long-distance number information is largely managed by two `number units'. Importantly, the behaviour of these units is partially controlled by other units independently shown to track syntactic structure. We conclude that LSTMs are, to some extent, implementing genuinely syntactic processing mechanisms, paving the way to a more general understanding of grammatical encoding in LSTMs." @default.
- W2952208026 created "2019-06-27" @default.
- W2952208026 creator A5007916691 @default.
- W2952208026 creator A5018069499 @default.
- W2952208026 creator A5028116010 @default.
- W2952208026 creator A5038612405 @default.
- W2952208026 creator A5059697591 @default.
- W2952208026 creator A5075892407 @default.
- W2952208026 date "2019-03-18" @default.
- W2952208026 modified "2023-10-02" @default.
- W2952208026 title "The emergence of number and syntax units in LSTM language models" @default.
- W2952208026 cites W1951216520 @default.
- W2952208026 cites W2037504148 @default.
- W2952208026 cites W2064675550 @default.
- W2952208026 cites W2141554956 @default.
- W2952208026 cites W2524611247 @default.
- W2952208026 cites W2529194139 @default.
- W2952208026 cites W2531381952 @default.
- W2952208026 cites W2549835527 @default.
- W2952208026 cites W2606347107 @default.
- W2952208026 cites W2606837722 @default.
- W2952208026 cites W2768794963 @default.
- W2952208026 cites W2798727047 @default.
- W2952208026 cites W2864832950 @default.
- W2952208026 cites W2888922637 @default.
- W2952208026 cites W2891343966 @default.
- W2952208026 cites W2892296496 @default.
- W2952208026 cites W2962776659 @default.
- W2952208026 cites W2962961857 @default.
- W2952208026 cites W2963430224 @default.
- W2952208026 cites W2963614302 @default.
- W2952208026 cites W2963751529 @default.
- W2952208026 cites W2964117978 @default.
- W2952208026 cites W2964159778 @default.
- W2952208026 cites W2964222268 @default.
- W2952208026 hasPublicationYear "2019" @default.
- W2952208026 type Work @default.
- W2952208026 sameAs 2952208026 @default.
- W2952208026 citedByCount "11" @default.
- W2952208026 countsByYear W29522080262019 @default.
- W2952208026 countsByYear W29522080262020 @default.
- W2952208026 countsByYear W29522080262021 @default.
- W2952208026 crossrefType "posted-content" @default.
- W2952208026 hasAuthorship W2952208026A5007916691 @default.
- W2952208026 hasAuthorship W2952208026A5018069499 @default.
- W2952208026 hasAuthorship W2952208026A5028116010 @default.
- W2952208026 hasAuthorship W2952208026A5038612405 @default.
- W2952208026 hasAuthorship W2952208026A5059697591 @default.
- W2952208026 hasAuthorship W2952208026A5075892407 @default.
- W2952208026 hasConcept C111919701 @default.
- W2952208026 hasConcept C125411270 @default.
- W2952208026 hasConcept C127705205 @default.
- W2952208026 hasConcept C137293760 @default.
- W2952208026 hasConcept C154945302 @default.
- W2952208026 hasConcept C15744967 @default.
- W2952208026 hasConcept C19417346 @default.
- W2952208026 hasConcept C204321447 @default.
- W2952208026 hasConcept C2775936607 @default.
- W2952208026 hasConcept C41008148 @default.
- W2952208026 hasConcept C60048249 @default.
- W2952208026 hasConceptScore W2952208026C111919701 @default.
- W2952208026 hasConceptScore W2952208026C125411270 @default.
- W2952208026 hasConceptScore W2952208026C127705205 @default.
- W2952208026 hasConceptScore W2952208026C137293760 @default.
- W2952208026 hasConceptScore W2952208026C154945302 @default.
- W2952208026 hasConceptScore W2952208026C15744967 @default.
- W2952208026 hasConceptScore W2952208026C19417346 @default.
- W2952208026 hasConceptScore W2952208026C204321447 @default.
- W2952208026 hasConceptScore W2952208026C2775936607 @default.
- W2952208026 hasConceptScore W2952208026C41008148 @default.
- W2952208026 hasConceptScore W2952208026C60048249 @default.
- W2952208026 hasOpenAccess W2952208026 @default.
- W2952208026 hasRelatedWork W1747312753 @default.
- W2952208026 hasRelatedWork W2064675550 @default.
- W2952208026 hasRelatedWork W2250473257 @default.
- W2952208026 hasRelatedWork W2485351821 @default.
- W2952208026 hasRelatedWork W2549835527 @default.
- W2952208026 hasRelatedWork W2741954161 @default.
- W2952208026 hasRelatedWork W2798727047 @default.
- W2952208026 hasRelatedWork W2918996109 @default.
- W2952208026 hasRelatedWork W2922523190 @default.
- W2952208026 hasRelatedWork W2943085977 @default.
- W2952208026 hasRelatedWork W2948124660 @default.
- W2952208026 hasRelatedWork W2948201212 @default.
- W2952208026 hasRelatedWork W2952349142 @default.
- W2952208026 hasRelatedWork W2962961857 @default.
- W2952208026 hasRelatedWork W2963430224 @default.
- W2952208026 hasRelatedWork W2963751529 @default.
- W2952208026 hasRelatedWork W2971628371 @default.
- W2952208026 hasRelatedWork W3016191525 @default.
- W2952208026 hasRelatedWork W3016825020 @default.
- W2952208026 hasRelatedWork W3172578930 @default.
- W2952208026 isParatext "false" @default.
- W2952208026 isRetracted "false" @default.
- W2952208026 magId "2952208026" @default.
- W2952208026 workType "article" @default.