Hugging Face Blog·· 3 h ago
ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
AI summary
ALTK-Evolve learns and applies a playbook using GPT-4.1, outperforming the retail entries on τ²-bench, a customer service agent benchmark—without updating model weights.
Selection record
Threshold 60Official, first-handFirst 58Second 58
Not admittedSum of both 116 < twice the threshold 120
- Source tier
- Official, first-hand; this tier's threshold is 60
- Pre-filter
- passed：介绍AI智能体记忆学习与基准评测方法
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Hugging Face Blog · huggingface.co