LLM agents (GPT-4o-mini, Gemini-2.5-flash, Claude-3.5-haiku) cannot learn an enumerated synthetic language through feedback within 100 turns, despite the task's grammar being fully specified in the system prompt.
On the cross-lingual transferability of monolingual representations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Talking with Oompa Loompas: A novel framework for evaluating linguistic acquisition of LLM agents
LLM agents (GPT-4o-mini, Gemini-2.5-flash, Claude-3.5-haiku) cannot learn an enumerated synthetic language through feedback within 100 turns, despite the task's grammar being fully specified in the system prompt.