3 ms·
Most benchmarks notoriously undercount multi-turn instruction following too. And it's where open source models (and Gemini) lose big time.
by ACCount37 3mo ago
Most benchmarks notoriously undercount multi-turn instruction following too. And it's where open source models (and Gemini) lose big time.