Preprint
Aug 2026
Actions Speak Louder than Words: Measuring Cross-Lingual Policy Retention in Tool-Using Agents
A single trace-extraction regex, not the model, manufactured a multilingual failure: two worked examples raise one model's measured accuracy twenty-sixfold while its accuracy on readable outputs barely moves.
Sourabrata Mukherjee, Kalika Bali, Sunayana Sitaram
· 0 citations