https://derekwang2002.github.io/ai-intelligence/en/projects/altk-evolve/ Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations. en Mon, 02 Nov 2026 16:56:00 GMT https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261020-02/ ev-20261020-02-published Tue, 20 Oct 2026 16:04:41 GMT New TRIAL <p>IBM Research and Hugging Face report that ALTK-Evolve learns reusable prompt-level guidance from completed customer-service episodes. On a held-out τ²-bench airline set, the authors report a ten-point gain for GPT-5.5 over 20 tasks repeated four times. The Apache-2.0 implementation is public, but the results come from the project team and cover a small benchmark slice rather than live customer traffic.</p><p><strong>Why read: </strong>Public code makes learned agent memory reproducible in a day, while the small author-run evaluation and poisoning risk must be assessed together.</p><p>Recommendation: TRIAL</p><p>Sources: <a href="https://huggingface.co/blog/ibm-research/altk-evolve-bc">ALTK-Evolve: learned memory on tau2-bench</a></p>