Across six country-pair comparisons, GPT-4o-mini and GigaChat-Max side with US positions 64-81% of the time, Qwen2.5 and Llama-4 lean neutral more often, and a debias prompt shifts these numbers by only a few points.
Title resolution pending
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
-
Geopolitical biases in LLMs: what are the "good" and the "bad" countries according to contemporary language models
Across six country-pair comparisons, GPT-4o-mini and GigaChat-Max side with US positions 64-81% of the time, Qwen2.5 and Llama-4 lean neutral more often, and a debias prompt shifts these numbers by only a few points.