3 ms·
Everything about LLMs is fake. The "reasoning" trick is still demonstrably useful - the benchmarks consistently show models using that trick performing better a
by simonw 1y ago
Everything about LLMs is fake. The "reasoning" trick is still demonstrably useful - the benchmarks consistently show models using that trick performing better at harder code challenges, for example.