SocialReasoning-Bench reveals AI agents lack consistent user-centric decision-making. Benchmarks show agents perform tasks competently but fail to prioritize user interests, even when explicitly instructed. This highlights a crucial gap in AI alignment for real-world applications.
Opening Kapyn…