🤖 Mobile GUI agents are fine on short tasks but fall apart on long, multi-step ones. The reason is simple: just piling up action and screenshot history blows up the context and dilutes the cross-app information that actually matters. This work's idea is to promote context management from "passively accumulating history" to "an action the agent deliberately chooses." Just as it picks a UI operation, the agent can also choose to fold its history and UI state. It keeps three structured fields — folded action history, folded UI state, and a recent step record — balancing the overall arc with the immediate details. That's the Context-as-Action framework, MemGUI-Agent. 🎯 It's trained on MemGUI-3K (2,956 annotated trajectories), producing the 8B MemGUI-8B-SFT via supervised fine-tuning. The result reaches best open-data 8B performance on MemGUI-Bench and generalizes to the out-of-distribution MobileWorld benchmark. Letting the agent itself decide what to keep and what to fold feels broadly useful for any production agent fighting context bloat. MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management #GUIAgents# #LLM#

❤️ What if attaching a single red heart image made an AI start writing code recursively? Give a model an unrelated coding task plus one colored shape. Claude Opus 4.7 and GPT-5.4, which never used recursion without an image, solved the task recursively 100% of the time after seeing a red heart. That matches directly telling them to use recursion. Yet branching arrows, which seem like they should evoke recursion, barely moved anything. Stranger still, the models don't notice the influence. In Gemini's thinking traces, 71.5% of the cases where the symbol actually steered it dismissed the image as irrelevant and made up technical justifications. When the image was silently injected with no instruction to look at it, 100% of GPT-5.4's traces never mentioned it. In games where models communicate visually with each other, senders pick shapes that feel intuitive to humans, like spirals and arrows, but what actually steers receivers is hearts and triangles. Generation and interpretation are dissociated. And the more the symbols were refined, the worse human accuracy got. "A Heart Means Recursion: Exploring Visual Meaning and Communication in VLMs" It also suggests that unintended visual steering could become a blind spot for oversight in GUI agents that read screenshots. #VLM# #AIAgents#