-
[Paper Review] SHIELD: An Auto-Healing Agentic Defense Framework for LLM Resource Exhaustion Attacks
Turning detection failures into knowledge — SHIELD pairs a three-stage defense pipeline with a knowledge-update and prompt-optimization loop to self-heal against sponge attacks.
-
[Paper Review] Code World Model Preparedness Report
Meta's pre-release frontier-risk assessment of CWM, a 32B open-weight code model, across cybersecurity, chemical & biological, and honesty (propensity) domains.
-
[Paper Review] A Survey on Generative Recommendation: Data, Model, and Tasks
A survey that reframes recommendation from discriminative scoring to generative synthesis, organized along the data, model, and task axes
-
[Paper Review] Multi-Interest Recommendation: A Survey
The first comprehensive survey of multi-interest recommendation, organizing the field around extractors, aggregators, diversity regularization, applications, and open challenges.
-
[Paper Review] Robust Uplift Modeling with Large-Scale Contexts for Real-time Marketing
A review of UMLC (KDD 2025): a model-agnostic uplift framework that tames the distribution shift and variance inflation caused by large-scale contexts via response-guided context grouping and feature interaction.
-
[Paper Review] Personalization and targeting: how to experiment, learn & optimize
An IJRM review that formalizes personalization as a causal-inference problem and walks through the test-and-learn cycle for experimenting, learning, and optimizing
-
[Paper Review] MobileExplorer: Accelerating On-Device Inference for Mobile GUI Agents via Online Exploration
Instead of letting slow on-device VLM inference idle, MobileExplorer spends that window probing the screen to gather hints for the next reasoning step.
-
[Paper Review] Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents
Apple's 3B on-device GUI agent. A close read of how real+synthetic data curation, zoom-in visual tool-use, and a two-stage SFT→RLVR recipe push a small model as far as it can go on GUI grounding and navigation — and where it honestly hits a wall.
-
[Paper Review] ShowUI: One Vision-Language-Action Model for GUI Visual Agent
A GUI agent that 'sees' screenshots like a human and clicks. A deep dive into ShowUI's UI-guided token selection and interleaved vision-language-action streaming, which hit 75.1% zero-shot grounding with a 2B model and 256K data.
-
[Paper Review] MinT: Managed Infrastructure for Training and Serving Millions of LLMs
A managed infrastructure for training and serving millions of LoRA policies over a small set of resident base models. Adapter revisions become the unit that crosses the training-serving boundary, cutting the handoff by 18.3x and validating the path up to 1T-scale MoE.
