Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385571421> ?p ?o ?g. }
Showing items 1 to 61 of
61
with 100 items per page.
- W4385571421 abstract "Recent work has shown that large pretrained Language Models (LMs) can not only perform remarkably well on a range of Natural Language Processing (NLP) tasks but also start improving on reasoning tasks such as arithmetic induction, symbolic manipulation, and commonsense reasoning with increasing size of models. However, it is still unclear what the underlying capabilities of these LMs are. Surprisingly, we find that these models have limitations on certain basic symbolic manipulation tasks such as copy, reverse, and addition. When the total number of symbols or repeating symbols increases, the model performance drops quickly. We investigate the potential causes behind this phenomenon and examine a set of possible methods, including explicit positional markers, fine-grained computation steps, and LMs with callable programs. Experimental results show that none of these techniques can solve the simplest addition induction problem completely. In the end, we introduce LMs with tutor, which demonstrates every single step of teaching. LMs with tutor is able to deliver 100% accuracy in situations of OOD and repeating symbols, shedding new insights on the boundary of large LMs in induction." @default.
- W4385571421 created "2023-08-05" @default.
- W4385571421 creator A5003805984 @default.
- W4385571421 creator A5033020983 @default.
- W4385571421 creator A5041666153 @default.
- W4385571421 creator A5047709762 @default.
- W4385571421 creator A5052596963 @default.
- W4385571421 date "2023-01-01" @default.
- W4385571421 modified "2023-09-24" @default.
- W4385571421 title "Limitations of Language Models in Arithmetic and Symbolic Induction" @default.
- W4385571421 doi "https://doi.org/10.18653/v1/2023.acl-long.516" @default.
- W4385571421 hasPublicationYear "2023" @default.
- W4385571421 type Work @default.
- W4385571421 citedByCount "0" @default.
- W4385571421 crossrefType "proceedings-article" @default.
- W4385571421 hasAuthorship W4385571421A5003805984 @default.
- W4385571421 hasAuthorship W4385571421A5033020983 @default.
- W4385571421 hasAuthorship W4385571421A5041666153 @default.
- W4385571421 hasAuthorship W4385571421A5047709762 @default.
- W4385571421 hasAuthorship W4385571421A5052596963 @default.
- W4385571421 hasBestOaLocation W43855714211 @default.
- W4385571421 hasConcept C154945302 @default.
- W4385571421 hasConcept C159985019 @default.
- W4385571421 hasConcept C177264268 @default.
- W4385571421 hasConcept C192562407 @default.
- W4385571421 hasConcept C199360897 @default.
- W4385571421 hasConcept C204323151 @default.
- W4385571421 hasConcept C2778371403 @default.
- W4385571421 hasConcept C33923547 @default.
- W4385571421 hasConcept C41008148 @default.
- W4385571421 hasConcept C45374587 @default.
- W4385571421 hasConcept C80444323 @default.
- W4385571421 hasConcept C94375191 @default.
- W4385571421 hasConceptScore W4385571421C154945302 @default.
- W4385571421 hasConceptScore W4385571421C159985019 @default.
- W4385571421 hasConceptScore W4385571421C177264268 @default.
- W4385571421 hasConceptScore W4385571421C192562407 @default.
- W4385571421 hasConceptScore W4385571421C199360897 @default.
- W4385571421 hasConceptScore W4385571421C204323151 @default.
- W4385571421 hasConceptScore W4385571421C2778371403 @default.
- W4385571421 hasConceptScore W4385571421C33923547 @default.
- W4385571421 hasConceptScore W4385571421C41008148 @default.
- W4385571421 hasConceptScore W4385571421C45374587 @default.
- W4385571421 hasConceptScore W4385571421C80444323 @default.
- W4385571421 hasConceptScore W4385571421C94375191 @default.
- W4385571421 hasLocation W43855714211 @default.
- W4385571421 hasOpenAccess W4385571421 @default.
- W4385571421 hasPrimaryLocation W43855714211 @default.
- W4385571421 hasRelatedWork W2013111119 @default.
- W4385571421 hasRelatedWork W2047793074 @default.
- W4385571421 hasRelatedWork W2112962394 @default.
- W4385571421 hasRelatedWork W2118300983 @default.
- W4385571421 hasRelatedWork W2382501300 @default.
- W4385571421 hasRelatedWork W3137189469 @default.
- W4385571421 hasRelatedWork W3162240892 @default.
- W4385571421 hasRelatedWork W4235530921 @default.
- W4385571421 hasRelatedWork W4243252198 @default.
- W4385571421 hasRelatedWork W4245713008 @default.
- W4385571421 isParatext "false" @default.
- W4385571421 isRetracted "false" @default.
- W4385571421 workType "article" @default.