---
type: "raw_signal"
id: "altk-evolve-how-learned-memory-tops-the-2-745e0f"
slug: "altk-evolve-how-learned-memory-tops-the-2-745e0f"
title: "ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model"
description: "IBM Research describes ALTK-Evolve, a method that distills behavioral guidelines from an agent’s training-task attempts and appends them to its system prompt without changing model weights."
retrieval_nugget: "IBM Research describes ALTK-Evolve, a method that distills behavioral guidelines from an agent’s training-task attempts and appends them to its system prompt without changing model weights."
observed_at: "2026-10-20"
record_date: "2026-10-20"
date_kind: "observed_at"
why_it_matters: "IBM Research describes ALTK-Evolve, a method that distills behavioral guidelines from an agent’s training-task attempts and appends them to its system prompt without changing model weights. In the authors’ τ²-bench tests, the full guideline set raised GPT-4.1 retail Pass⁴ from 57.5% to 67.5% on held-out tasks. In airline tasks using GPT-5.5, 103 guidelines raised Pass⁴ from 55% to 70%. The results"
novelty: "notable"
verification_level: "source-linked"
signal_type: "research"
source_platform: "huggingface.co"
topics: []
entities: ["IBM Research","ALTK-Evolve","τ²-bench","GPT-4.1","GPT-5.5"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "Adds a customer-service benchmark result: ALTK-Evolve with GPT-4.1 scored 67.5% Pass^4 on τ²-bench retail, topping the October 12 leaderboard snapshot without model-weight updates."
prior_refs: ["memory:c7968e19-d303-8a00-ab25-4af0fb99cbce","memory:ecc5d3fb-ef40-8e1c-a19e-ccebebcee1cf"]
source_url: "https://huggingface.co/blog/ibm-research/altk-evolve-bc"
source_urls: ["https://huggingface.co/blog/ibm-research/altk-evolve-bc"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:altk-evolve-how-learned-memory-tops-the-2-745e0f"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/signals/altk-evolve-how-learned-memory-tops-the-2-745e0f/","markdown":"https://newruntime.com/signals/altk-evolve-how-learned-memory-tops-the-2-745e0f.md","json":"https://newruntime.com/signals/altk-evolve-how-learned-memory-tops-the-2-745e0f.json"}
---

# ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model

## Retrieval answer

IBM Research describes ALTK-Evolve, a method that distills behavioral guidelines from an agent’s training-task attempts and appends them to its system prompt without changing model weights.

IBM Research describes ALTK-Evolve, a method that distills behavioral guidelines from an agent’s training-task attempts and appends them to its system prompt without changing model weights. In the authors’ τ²-bench tests, the full guideline set raised GPT-4.1 retail Pass⁴ from 57.5% to 67.5% on held-out tasks. In airline tasks using GPT-5.5, 103 guidelines raised Pass⁴ from 55% to 70%. The results are benchmark-specific: τ²-bench uses simulated customers, documented policies, and tool actions, and the reported test sets are small. The article also reports that the best guideline volume varies by domain, and that gains come with more prompt tokens rather than longer conversations.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)