PingPong: Handcrafted Code-Switching Dialogue Benchmark Accepted by EMNLP 2026

SARAL's PingPong benchmark, accepted by EMNLP 2026, is a rare hand-annotated dataset for code-switched multi-party dialogue covering 6 language pairs. The paper shows current LLMs still struggle with natural code-switching.

2026-09-14 ~ 2026-09-14 · 2 related posts