Bookmarks
A note about "The Humane Representation of Thought"
A year and a half ago, on a plane, I wrote An Ill-Advised Personal Note about "Media for Thinking the Unthinkable".
Working memory - Wikipedia
Working memory is a cognitive system with a limited capacity that can hold information temporarily. [1] It is important for reasoning and the guidance of decision-making and behavior.
Death Note: L, Anonymity & Eluding Entropy
The text discusses Light's mistakes in using the Death Note and how they led to his de-anonymization by L. Light's errors, such as revealing his precise killing methods and using confidential police information, significantly reduced his anonymity. The text also explores strategies Light could have employed to better protect his anonymity while using the Death Note.
Generative Agents: Interactive Simulacra of Human Behavior
The content discusses generative agents that simulate believable human behavior for interactive applications. These agents populate a sandbox environment, interact with each other, plan their days, form relationships, and exhibit emergent social behaviors. The paper introduces a novel architecture that allows agents to remember, retrieve, reflect, and interact dynamically.
K-Level Reasoning with Large Language Models
Large Language Models (LLMs) have shown proficiency in complex reasoning tasks, but their performance in dynamic and competitive scenarios remains unexplored. To address this, researchers have introduced two game theory-based challenges that mirror real-world decision-making. Existing reasoning methods tend to struggle in dynamic settings that require k-level thinking, so the researchers propose a novel approach called "K-Level Reasoning" that improves prediction accuracy and informs strategic decision-making. This research sets a benchmark for dynamic reasoning assessment and enhances the proficiency of LLMs in dynamic contexts.
Measuring Faithfulness in Chain-of-Thought Reasoning
Large language models (LLMs) are more effective when they engage in step-by-step "Chain-of-Thought" (CoT) reasoning, but it is unclear if this reasoning is a faithful explanation of the model's actual process. The study examines how interventions on the CoT affect model predictions, finding that models vary in how strongly they rely on the CoT. The performance boost from CoT does not solely come from added test-time compute or specific phrasing. As models become larger and more capable, they tend to produce less faithful reasoning. The results suggest that faithful CoT reasoning depends on carefully chosen circumstances such as model size and task.
Subcategories
- applications (15)
- computer_architecture (1)
- ethics (1)
- expert_systems (2)
- game_ai (5)
- knowledge_representation (4)
- machine_learning (324)
- natural_language_processing (3)
- planning_and_scheduling (2)
- robotics (2)
- software_development (1)
- theory (1)