The Morning
From the arXiv
ECHO: A Locally-Deployable Agentic Health Assistant with Temporal Memory, Safety Guardrails, and Speech Assessment
CHO is a locally-deployable health assistant for chronic care that uses an agentic chatbot with temporal memory and clinical tools. Its core contribution lies in its robust safety system, combining a fast rule-based layer with a GNN for accurate clinical intent classification, ensuring secure and effective long-term health management.
Learning Globally Reusable Skills for Coding Agents
This paper introduces GSE, a framework for improving LLM coding agents by evolving skills globally. Unlike local updates, GSE uses a Skill Relation Graph to model inter-skill relationships and cluster-based consolidation to abstract reusable capabilities, enha…
TRAJDEBUG: Tracing Error Lifecycle to Identify Critical Failures in Long-Horizon Agent Trajectories
TrajDebug is a framework designed to pinpoint the root cause of failures in long, complex trajectories generated by LLM-based agents. It addresses challenges of scattered evidence and multiple local errors by using multi-granularity history compression and evi…
A Six-Dimensional Taxonomy of Post-Training Adaptation Techniques with Applications in AI Governance
This paper introduces a six-dimensional taxonomy to systematically categorize and understand the diverse landscape of post-training adaptation techniques in machine learning. Its core contribution lies in providing a unified framework that clarifies distinctio…
The Bitter Lesson of Tool Calling
This paper introduces Programmatic Tool Calling (PTC), a method that allows LLMs to use tools by invoking them as typed Python scripts, enabling natural chaining and parallelization. Empirically, PTC matches or surpasses traditional JSON tool calling across mo…
Training-Free Token-Level Steering for LLM Personalized Co-Writing
This paper introduces SteerWrite, a training-free framework for personalized LLM co-writing. Its core method achieves token-level steering without gradient updates, making it effic…
Benchmarking the Benchmarks: Evaluating Benchmarks for Conversational Agents
This paper introduces a novel, reference-free framework that leverages LLM judges to evaluate the quality of conversational agent benchmarks. The core method assesses benchmark con…
Contextual Information Policy Optimization for Search Agents
This paper introduces Contextual Information Policy Optimization (CIPO), a novel reinforcement learning framework for search agents. CIPO addresses the problem of agents relying to…
DASH: Divergence-Adaptive Supervision Horizons for On-Policy Self-Distillation of Reasoning Models
This paper introduces DASH, a novel on-policy self-distillation method for improving reasoning in language models. DASH addresses the issue of sparse rewards in reinforcement learn…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
EnvACE trains large language model agents for complex tool use by replacing costly environment interactions with "world rehearsal." The agent learns by alternating between generati…
The Town Square
AMD acquired Taalas to integrate AI models directly into silicon, aiming to significantly enhance inference performance.
Workshops
This repository provides production-grade engineering skills for AI coding agents, enabling them to perform complex tasks like code generation, debugging, and testing.
This repository provides a framework for giving AI agents access to a simulated computer environment, enabling them to execute code, interact with files, and perform complex tasks.
