The Morning
From the arXiv
Extending LLM Context via Associative Recurrent Memory
his paper introduces the Associative Recurrent Memory Transformer (ARMT) to overcome the context length limitations of standard transformers. ARMT achieves constant memory scaling and improved efficiency by incorporating an associative recurrent memory mechanism. The authors also contribute domain-specific long-context datasets and a training recipe that combines pre-training, synthetic data, and curriculum learning to effectively extend LLM context.
Proxy Exploration and Reusable Guidance: A Modular LLM Post-Training Paradigm via Proxy-Guided Update Signals
This paper introduces Proxy-guided Update Signal Transfer (PUST), a modular post-training framework for LLMs. PUST decouples exploration from alignment by using a lightweight proxy model to efficiently discover high-reward behaviors. The relative improvement s…
Xiaomi-Robotics-U0: Unified Embodied Synthesis with World Foundation Model
This paper introduces Xiaomi-Robotics-U0, a unified 38-billion-parameter model for embodied AI. Its core method is to treat embodied generation as an extension of existing foundation image/video models, jointly optimizing various tasks like scene and video gen…
HyperSafe: Inference-Time Safety Recovery for Fine-Tuned Language Models
HyperSafe addresses the fragility of safety in fine-tuned language models by proposing a post-hoc, model-specific approach. It generates a "Safe Side Network" (SSN) for each fine-tuned model by learning a mapping from layer-wise activation fingerprints to SSN …
GEIS: A Generation-Evaluation-Improvement Loop of Agent Skills for Long-Form Article Generation
GEIS introduces a novel Generation-Evaluation-Improvement loop for long-form article generation. Its core method involves a structured cycle of agent skills, including planning, drafting, and evaluation, to address challenges like long context and complex inst…
Agent Hacks Agent: Autoresearch for Production-Agent Red-Teaming
This paper introduces AHA, an automated red-teaming method for LLM agents. AHA uses one agent to discover vulnerabilities in another by forming and testing hypotheses about unsafe …
From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligence
This paper proposes a roadmap for developing physical intelligence by addressing fragmentation in current research. It identifies key gaps in model representation, standardization,…
MM-ToolSandBox: A Unified Framework for Evaluating Visual Tool-Calling Agents
MM-ToolSandBox is a unified framework and benchmark designed to evaluate visual tool-calling agents. It offers a stateful execution environment with over 500 tools across diverse d…
RAGU: A Multi-Step GraphRAG Engine with a Compact Domain-Adapted LLM
RAGU is a modular GraphRAG engine that improves knowledge graph construction by separating extraction from consolidation, using a two-stage process with deduplication and summariza…
When Local Monitors Miss Compositional Harm: Diagnosing Distributed Backdoors in Multi-Agent Systems
This paper introduces the concept of "distributed backdoors" in multi-agent LLM systems, where a harmful payload is split across agents such that individual local checks pass. The …
The Town Square
Samsung Health users face potential data deletion if they decline AI training, raising privacy concerns.
Workshops
The Destructive Command Guard (dcg) is a tool designed to prevent dangerous Git and shell commands from being executed by agents, thereby safeguarding systems from accidental or malicious data loss.
Vibe-Trading is a personal trading agent that automates trading strategies, offering features for strategy development, backtesting, and live execution.
