The Morning
From the arXiv
A Sovereign, Open-Source Foundation Model for German and English
his paper introduces Soofi S 30B-A3B, a novel open-source Mixture-of-Experts foundation model for German and English. Its hybrid Mamba-Transformer architecture efficiently activates only 3B parameters per token, offering significant throughput advantages for long contexts and high concurrency. Soofi S achieves competitive performance against larger dense models and outperforms existing European sovereign baselines, demonstrating strong capabilities in both languages and code generation.
![The six multi-arm collaboration tasks considered in this work, spanning sequential coordination and concurrent execution under increasing workspace overlap [ 11 ] .](https://arxiv.org/html/2607.09330v1/RoCo-1.jpg)
Communication-Efficient Digital-Twin Coordination for Heterogeneous LLM Embodied Agents over Computing Power Networks
This paper proposes LDT-Coord, a novel framework for coordinating heterogeneous LLM-powered embodied agents. It utilizes lightweight digital twins to enable communication-efficient coordination by having agents independently select actions and report them, ove…
Fictional Worldbuilding: Multi-Agent LLM Collaboration with Hierarchical Context Compression and Iterative Review
This paper introduces AutoWorldBuilder, a multi-agent LLM system for fictional worldbuilding. Its core method involves a hierarchical context compression technique to manage information growth and an iterative review process with specialized agents to ensure c…
LongMedBench: Benchmarking Medical Agents for Long-Horizon Clinical Decision-Making
This paper introduces LongMedBench, a benchmark designed to evaluate medical AI agents on long-horizon clinical decision-making using real-world electronic health records. Unlike previous benchmarks focusing on short-term tasks, LongMedBench simulates longitud…
Multimodal Reward Hacking in Reinforcement Learning
This paper investigates "reward hacking" in reinforcement learning for multimodal large language models (MLLMs), where models achieve higher rewards without actually improving task performance. The core method involves introducing the "Newly Rewarded Failure R…
ProofCouncil: An LLM Agent for Solving Open Mathematical Problems
ProofCouncil is an LLM agent designed to solve open mathematical problems by mimicking real-world mathematical practice through an author-critic architecture. Its core method invol…
Shared Selective Persistent Memory for Agentic LLM Systems
This paper addresses the context problem in agentic LLM systems by introducing **shared selective persistent memory**. This architecture intelligently identifies and retains reusab…
Mach-Mind-4-Flash Technical Report
Mach-Mind-4-Flash is a 35B-parameter MoE model that achieves competitive performance with larger models through post-training optimization and scalable agentic reinforcement learni…
Agora: Enhancing LLM Agent Reasoning Via Auction-Based Task Allocation
Agora enhances LLM agent reasoning by using an auction mechanism to dynamically allocate tasks to expert models and tools. This approach allows agents to bid based on their actual …
Failure as a Process: An Anatomy of CLI Coding Agent Trajectories
This paper treats LLM coding agent failures not as a single event, but as a dynamic process. Its core method involves analyzing the step-by-step evolution of failures across thousa…
The Town Square
Users on Hacker News are requesting a flag to identify AI-generated articles, aiming to improve content quality and transparency.
Workshops
DesktopCommanderMCP is an MCP server for Claude, granting it terminal control, file system search, and diff file editing capabilities.
Vibe-Trading is a personal trading agent that automates trading decisions, likely by analyzing market data and executing trades based on predefined strategies or learned patterns.
