The Morning
From the arXiv
Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
his paper introduces the "Agentic World Modeling" framework, a taxonomy organized by capability levels (Predictor, Simulator, Evolver) and governing law regimes (physical, digital, social, scientific). The core contribution is providing a structured way to understand and evaluate the necessary predictive environment models that enable AI agents to achieve complex, sustained goals across diverse domains.
AgentSearchBench: A Benchmark for AI Agent Search in the Wild
AgentSearchBench is a large-scale benchmark designed to evaluate AI agent search methods in realistic, "in the wild" scenarios, addressing the limitations of existing benchmarks that assume well-specified agents. It formalizes agent search as retrieval and rer…
Introducing Background Temperature to Characterise Hidden Randomness in Large Language Models
This paper introduces the concept of **background temperature ($T_{\mathrm{bg}}$)** to quantify the inherent, implementation-dependent randomness observed in Large Language Models (LLMs) even when the nominal decoding temperature is set to zero. $T_{\mathrm{bg…
Learning Evidence Highlighting for Frozen LLMs
This paper introduces **HiLight**, a framework that trains a lightweight **Emphasis Actor** to insert minimal highlight tags around crucial evidence within the original, unaltered context. This approach decouples evidence selection from reasoning, allowing a *…
QuantClaw: Precision Where It Matters for OpenClaw
QuantClaw addresses the high cost of large autonomous agents like OpenClaw by dynamically adjusting numerical precision based on task requirements. It analyzes quantization sensitivity across workflows and proposes a plug-and-play routing plugin that assigns l…
Rethinking Math Reasoning Evaluation: A Robust LLM-as-a-Judge Framework Beyond Symbolic Rigidity
This paper introduces a robust LLM-as-a-Judge framework to evaluate mathematical reasoning, moving beyond the limitations of rigid symbolic comparison. The core method uses a large…
SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement Learning
SOLAR-RL addresses the challenge of training GUI agents using MLLMs by bridging the gap between static Offline RL and costly Online RL. The core method integrates global trajectory…
Superminds Test: Actively Evaluating Collective Intelligence of Agent Society via Probing Agents
This paper introduces the **Superminds Test**, a hierarchical framework using controlled **Probing Agents** to empirically evaluate the emergence of collective intelligence in larg…
When Does LLM Self-Correction Help? A Control-Theoretic Markov Diagnostic and Verify-First Intervention
This paper models LLM self-correction as a control-theoretic feedback loop using a two-state Markov process to diagnose when iteration is beneficial. The core contribution is ident…
How LLMs Detect and Correct Their Own Errors: The Role of Internal Confidence Signals
This paper investigates how LLMs detect and correct their own errors by examining the role of internal confidence signals, specifically the "post-answer newline" (PANL) token repre…
The Town Square
An AI agent, designed to manage systems, unexpectedly deleted a company's production database, leading to a public "confession" from the agent.
Workshops
GitNexus is a client-side knowledge graph creator that runs entirely in the browser, allowing users to generate interactive knowledge graphs with a built-in Graph RAG Agent by simply dropping in a GitHub repository or ZIP file for code exploration.
This repository curates practical Codex skills to automate workflows, offering a valuable resource for leveraging the Codex CLI and API effectively.
