ALTK-Evolve tops τ²-bench customer service leaderboard without a frontier model
Read original source →huggingface.coIBM Research's ALTK-Evolve reaches the top of the τ²-bench customer service leaderboard using learned memory rather than a frontier model. The Hugging Face post details how the memory approach outperforms larger models on the benchmark.
1 source
More stories today
Higgsfield AI compares MiniMax M3.1 Flash and Claude Opus 5.5
Higgsfield AI 🧩·40 minutes ago
ElevenLabs launches Eleven Realtime voice-agent models Mini and Max
Eleven Realtime bundles speech recognition, reasoning, turn-taking, and voice generation into two managed models, Mini and Max. Both use a cascaded architecture, but developers can't select or independently configure the underlying components.
The New Stack·53 minutes ago
ChatGPT adds PIN- and biometric-locked chats
Locked Chats are secured with a four-digit PIN or Face ID/Touch ID and move to a separate section hidden from chat history, pinned chats, and search. OpenAI says locking doesn't add end-to-end encryption, change retention, or alter the "Improve the model for everyone" setting.
TechCrunch·1 hour ago
OpenAI acquires Orchestra Bio for life sciences R&D
OpenAI is acquiring Orchestra Bio to bring AI-powered operations to life sciences R&D, aiming to help biopharma teams accelerate drug development. No financial terms were disclosed.
OpenAI Blog·1 hour ago
New multilingual dataset targets a linguistic phenomenon, with n-gram baselines
Anna Rogers·1 hour agoNVIDIA Cosmos3-H-Dreams generates clinical surgery simulations
NVIDIA's Cosmos3-H-Dreams model is presented as a generative simulation system for clinical surgery, per a Hugging Face blog post. The post frames it as bringing surgical scenarios to life through synthetic video rather than real footage.
Hugging Face Blog·1 hour ago
Databricks introduces MemCurrent persistent memory for production agents
MemCurrent combines episodic memory with derived memory across three access patterns: plugins that log and load context automatically, tools the agent calls, and a third pattern. Databricks frames it as a fix for production agents that "start from zero" each session.
Databricks Blog·1 hour ago
Franchisee cuts menu repricing time 95% with ChatGPT Work
An independent Schlotzsky's and Cinnabon franchisee cut menu repricing time by 95% and reconciles cash in 94% less time using ChatGPT Work. The workflow could free up $100K in cash flow.
