Papers, findings, and ideas that will define what AI becomes next.
Researchers have developed a novel method for extracting lithium, a critical component for electric vehicle batteries. This new technique offers the potential to be significantly more environmentally friendly and cost-effective than current industrial methods.
A new method for extracting lithium could significantly cut costs and emissions associated with producing lithium, a critical material for electric vehicles.
This research introduces a Q-Q orthogonality formulation to separate the causes of instability in sample quantiles derived from heavy-tailed distributions, specifically addressing the effects of projection direction and quantile thresholds.
This research audits the use of ensemble disagreement as a proxy for uncertainty in medical image segmentation, finding that Deep Ensembles (DE) provide superior calibration and failure detection compared to standard K-fold Cross-Validation (CV) ensembles. The study suggests sele
This research investigates how features are learned in two-layer neural networks under linear-width constraints, comparing single-step versus multi-step gradient descent. It demonstrates that reusing batches allows the model to capture multiple learned directions, extending the b
Researchers introduce MELT, the first behavioral trace dataset designed to analyze and detect high-risk memecoin launches on Solana. The dataset parses massive blockchain activity into typed behavioral records, revealing coordination strategies and ownership concealment among coordinators
Researchers introduce GENSTRAT, a novel methodology that uses procedurally generated strategic game environments to evaluate Large Language Models (LLMs) as economic agents. This approach provides deployment-relevant diagnostics by assessing model competence across six strategic.
Introducing the Foundation Protocol (FP), a graph-first coordination layer designed to manage the complex interactions, economic primitives, and accountability required by increasingly autonomous AI agents operating in the real world.
This research introduces 'AutoResearch,' a framework defining the spectrum of AI-powered scientific workflow automation. It analyzes the transition from task-level assistance to full research automation, identifying critical gaps in reproducibility, provenance, and accountability
This paper introduces an effective hybrid approach combining Dynamic Programming (DP) as the primary search framework and Constraint Programming (CP) for global constraint propagation to solve the complex Partial Shop Scheduling Problem (PSSP).
A systematic study investigates the role of context, external moral knowledge (retrieval), and model scaling in detecting Schwartz values within political texts. The research finds that context and retrieval are most beneficial for socially situated values, suggesting that context
This research introduces a capability-level theory for placing accountability boundaries within agentic AI systems, introducing 'accountability assets' and 'rule debt' to address how responsibility transfers as AI capabilities become modular and distributed.
Introduces ProxySHAP, a novel method that uses tree-based proxy models and residual correction to efficiently approximate complex Shapley and Banzhaf interactions, establishing a new state-of-the-art for model explainability.
A new method, Inductive Deductive Synthesis (IDS), allows AI agents to jointly synthesize implementations and formal proofs, dramatically accelerating the generation of formally verified systems compared to expert human effort.
New research establishes fundamental, architecture-based limits on the performance and feasibility of various AI tasks (like preference learning and verification). The work translates these theoretical impossibility results into concrete design specifications for building truly trustworthy AI.
This paper introduces Ontological Knowledge Blocks (OKBs), a governance infrastructure designed to automate compliance and validation for AI systems deployed in critical infrastructure. OKBs compile regulatory obligations into machine-checkable constraints, enabling profile-based
A new framework, EVE-Agent, is introduced to address the issue of unreliable training signals in self-evolving search agents. By requiring all generated examples to include verifiable evidence spans, the agent can learn from evidence that genuinely improves correctness.
Hybrid search strategies are critical for transitioning Retrieval-Augmented Generation (RAG) systems from prototypes to production-ready solutions.
This research introduces Mediative Fuzzy Logic, a new framework designed to reconcile conflicting assessments in fuzzy control and decision-making. It develops a unified mathematical account extending standard fuzzy logic through interval type-2, granular type-3, and quantum extensions
This paper proposes a three-step framework for designing benchmarks for LLM agents engaged in knowledge work. It addresses the gap where traditional NLP evaluations fail to measure real-world competence, focusing on explicitly defining work activities, testing settings, and scoring.
Explore the journey of semantic search, tracing its evolution from simple keyword matching (TF-IDF) to modern, transformer-based language understanding. A hands-on guide detailing the implementation of four generations of semantic search systems using Python.
Pope Leo XIV's inaugural encyclical employs Artificial Intelligence not as a subject, but as a framework to analyze systemic societal issues, focusing on concentrated power and the influence of the tech elite.
A recent study explores the capabilities of AI tools like ChatGPT when assisting with coding tasks related to causal inference using statistical languages such as Python, R, and Stata.
Nous Research introduces Contrastive Neuron Attribution (CNA), a novel method for steering Large Language Model (LLM) behavior by identifying and ablating sparse MLP neuron circuits without requiring sparse autoencoder training or weight modification.