A language-only question-answering reference game gives language models a communicative reward signal, but fine-tuning with it changes speaker behavior without improving grammatical ability.
Child's Talk : Learning to Use Language
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Towards Developmentally Plausible Rewards: Communicative Success as a Learning Signal for Interactive Language Models
A language-only question-answering reference game gives language models a communicative reward signal, but fine-tuning with it changes speaker behavior without improving grammatical ability.