IBM Research's ALTK-Evolve, a system using learned memory, reached the top of the τ²-bench leaderboard on customer service tasks without relying on a frontier model, according to a post on Hugging Face1. The approach suggests competitive benchmark performance through memory-based techniques rather than scaling to frontier-class architectures.
IBM's ALTK-Evolve Tops τ²-Bench Customer Service Leaderboard Without Frontier Model
IBM Research's ALTK-Evolve reached the top of the τ²-bench leaderboard on customer service tasks using learned memory, without a frontier model.