Memory Injections: Correcting Multi-Hop Reasoning Failures During Inference in Transformer-Based Language Models.
Mansi SakarvadiaAswathy AjithArham KhanDaniel GrzendaNathaniel HudsonAndré BauerKyle ChardIan T. FosterPublished in: BlackboxNLP@EMNLP (2023)
Keyphrases
- language model
- multi hop
- language modeling
- wireless networks
- wireless sensor networks
- data transmission
- ad hoc networks
- mobile ad hoc networks
- end to end
- n gram
- base station
- probabilistic model
- language modelling
- retrieval model
- document retrieval
- routing protocol
- energy efficient
- speech recognition
- smoothing methods
- information retrieval
- energy consumption
- query expansion
- relevance model
- routing algorithm
- statistical language models
- context sensitive
- test collection
- wireless communication
- language models for information retrieval
- wifi
- pseudo relevance feedback
- communication networks
- document ranking
- bayesian networks