
PACT narrows plan‑action gaps in GUI agents, boosting mobile and desktop task success
Researchers introduce PACT, a training framework that preserves correct plans and repairs actions in GUI agents.
The AI news that matters for people who make things — what happened, and how to put it to work in Grafy.

Researchers introduce PACT, a training framework that preserves correct plans and repairs actions in GUI agents.

RadixArk released DaisyRose-27B, a Qwen‑based model that doubles Terminal‑Bench pass@1 while cutting training cost by 57%.

Researchers introduce IBMARL, a multi‑agent reinforcement learning framework that leverages single‑human demonstrations and outperforms prior demo‑based.

Researchers unveil ARCHER, a two‑stage framework that turns free‑form rationales into executable traces, raising small model math accuracy by up to 9 points.

A recent paper finds that decision models that struggle as actors can still serve as effective teachers for reinforcement‑learning agents.

A controlled RLVR study finds shared parameters suppress narrowly shared reward hacks and amplify widely shared ones, and shows what to audit.

Researchers found that round‑to‑nearest bf16 conversion silently drops most policy updates during RL fine‑tuning.