Hugging Face Blog·· 2 hours agoFeaturedAI Score60
ALTK-Evolve: learned prompt guidelines take a GPT-4.1 agent to the top of τ²-bench retail without weight updates
ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
AI summary
ALTK-Evolve uses GPT-4.1 to distill an agent's own τ²-bench training attempts into short behavioral guidelines appended to the system prompt.
Why it matters
The results show prompt-level learned guidelines lifting Pass^4 reliability without weight updates, and how the right guideline count differs by domain.
Source: Hugging Face Blog · huggingface.co