FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3085932930> ?p ?o ?g. }

• W3085932930 abstract "Current approaches to text generation largely rely on autoregressive models and maximum likelihood estimation. This paradigm leads to (i) diverse but low-quality samples due to mismatched learning objective and evaluation metric (likelihood vs. quality) and (ii) exposure bias due to mismatched history distributions (gold vs. model-generated). To alleviate these problems, we frame text generation as a reinforcement learning (RL) problem with expert demonstrations (i.e., the training data), where the goal is to maximize quality given model-generated histories. Prior RL approaches to generation often face optimization issues due to the large action space and sparse reward. We propose GOLD (generation by off-policy learning from demonstrations): an algorithm that learns from the off-policy demonstrations by importance weighting and does not suffer from degenerative solutions. We find that GOLD outperforms the baselines according to automatic and human evaluation on summarization, question generation, and machine translation, including attaining state-of-the-art results for CNN/DailyMail summarization. Further, we show that models trained by GOLD are less sensitive to decoding algorithms and the generation quality does not degrade much as the length increases." @default.
• W3085932930 created "2020-09-21" @default.
• W3085932930 creator A5058459753 @default.
• W3085932930 creator A5086654042 @default.
• W3085932930 date "2020-09-16" @default.
• W3085932930 modified "2023-09-27" @default.
• W3085932930 title "Text Generation by Learning from Off-Policy Demonstrations." @default.
• W3085932930 cites W1514587017 @default.
• W3085932930 cites W1544827683 @default.
• W3085932930 cites W1771410628 @default.
• W3085932930 cites W2064675550 @default.
• W3085932930 cites W2101105183 @default.
• W3085932930 cites W2119717200 @default.
• W3085932930 cites W2138309709 @default.
• W3085932930 cites W2154652894 @default.
• W3085932930 cites W2155027007 @default.
• W3085932930 cites W2157331557 @default.
• W3085932930 cites W2158349948 @default.
• W3085932930 cites W2606974598 @default.
• W3085932930 cites W2797269287 @default.
• W3085932930 cites W2888482885 @default.
• W3085932930 cites W2898718449 @default.
• W3085932930 cites W2962935506 @default.
• W3085932930 cites W2962953307 @default.
• W3085932930 cites W2962957031 @default.
• W3085932930 cites W2963248296 @default.
• W3085932930 cites W2963382396 @default.
• W3085932930 cites W2963403868 @default.
• W3085932930 cites W2963466651 @default.
• W3085932930 cites W2963620441 @default.
• W3085932930 cites W2963748441 @default.
• W3085932930 cites W2963899977 @default.
• W3085932930 cites W2964268978 @default.
• W3085932930 cites W2968831808 @default.
• W3085932930 cites W2970692082 @default.
• W3085932930 cites W2970717545 @default.
• W3085932930 cites W2982399380 @default.
• W3085932930 cites W2995338136 @default.
• W3085932930 cites W2995404354 @default.
• W3085932930 cites W2996068536 @default.
• W3085932930 cites W2996287690 @default.
• W3085932930 cites W3022566517 @default.
• W3085932930 cites W3035214886 @default.
• W3085932930 hasPublicationYear "2020" @default.
• W3085932930 type Work @default.
• W3085932930 sameAs 3085932930 @default.
• W3085932930 citedByCount "5" @default.
• W3085932930 countsByYear W30859329302020 @default.
• W3085932930 countsByYear W30859329302021 @default.
• W3085932930 crossrefType "posted-content" @default.
• W3085932930 hasAuthorship W3085932930A5058459753 @default.
• W3085932930 hasAuthorship W3085932930A5086654042 @default.
• W3085932930 hasConcept C111472728 @default.
• W3085932930 hasConcept C11413529 @default.
• W3085932930 hasConcept C119857082 @default.
• W3085932930 hasConcept C126838900 @default.
• W3085932930 hasConcept C138885662 @default.
• W3085932930 hasConcept C144024400 @default.
• W3085932930 hasConcept C149782125 @default.
• W3085932930 hasConcept C154945302 @default.
• W3085932930 hasConcept C159877910 @default.
• W3085932930 hasConcept C162324750 @default.
• W3085932930 hasConcept C170858558 @default.
• W3085932930 hasConcept C176217482 @default.
• W3085932930 hasConcept C183115368 @default.
• W3085932930 hasConcept C21547014 @default.
• W3085932930 hasConcept C2779304628 @default.
• W3085932930 hasConcept C2779530757 @default.
• W3085932930 hasConcept C33923547 @default.
• W3085932930 hasConcept C36289849 @default.
• W3085932930 hasConcept C41008148 @default.
• W3085932930 hasConcept C57273362 @default.
• W3085932930 hasConcept C71924100 @default.
• W3085932930 hasConcept C97541855 @default.
• W3085932930 hasConceptScore W3085932930C111472728 @default.
• W3085932930 hasConceptScore W3085932930C11413529 @default.
• W3085932930 hasConceptScore W3085932930C119857082 @default.
• W3085932930 hasConceptScore W3085932930C126838900 @default.
• W3085932930 hasConceptScore W3085932930C138885662 @default.
• W3085932930 hasConceptScore W3085932930C144024400 @default.
• W3085932930 hasConceptScore W3085932930C149782125 @default.
• W3085932930 hasConceptScore W3085932930C154945302 @default.
• W3085932930 hasConceptScore W3085932930C159877910 @default.
• W3085932930 hasConceptScore W3085932930C162324750 @default.
• W3085932930 hasConceptScore W3085932930C170858558 @default.
• W3085932930 hasConceptScore W3085932930C176217482 @default.
• W3085932930 hasConceptScore W3085932930C183115368 @default.
• W3085932930 hasConceptScore W3085932930C21547014 @default.
• W3085932930 hasConceptScore W3085932930C2779304628 @default.
• W3085932930 hasConceptScore W3085932930C2779530757 @default.
• W3085932930 hasConceptScore W3085932930C33923547 @default.
• W3085932930 hasConceptScore W3085932930C36289849 @default.
• W3085932930 hasConceptScore W3085932930C41008148 @default.
• W3085932930 hasConceptScore W3085932930C57273362 @default.
• W3085932930 hasConceptScore W3085932930C71924100 @default.
• W3085932930 hasConceptScore W3085932930C97541855 @default.
• W3085932930 hasLocation W30859329301 @default.
• W3085932930 hasOpenAccess W3085932930 @default.
• W3085932930 hasPrimaryLocation W30859329301 @default.
• W3085932930 hasRelatedWork W133075886 @default.

• next