Science – usdaed

Beyond Algorithms: Building Intelligent Agents for Finance

03.12.2025 by qfx

Agentic trading is framed as an orchestration of specialized agents-including a contextual memory agent-that maps onto components within the FinAgent framework, suggesting a system designed to leverage nuanced understanding beyond purely algorithmic approaches.

A new framework is emerging that moves beyond traditional algorithmic trading to create fully autonomous, language-driven agents capable of navigating complex financial markets.

Decoding the Market with AI: Finding Hidden Connections in Prediction Trades

03.12.2025 by qfx

The agent demonstrates category-specific relational prediction accuracy, achieving varying levels of performance across different categories as indicated by aggregated results across all clusters within each label.

New research demonstrates how artificial intelligence can analyze prediction markets to uncover relationships between different contracts, potentially leading to better forecasts and trading strategies.

NFT Scams: Following the Money on the Blockchain

03.12.2025 by qfx

The study dissects the mechanics of NFT airdrop phishing, exposing how malicious actors leverage the promise of free digital assets to initiate attacks and compromise unsuspecting users.

A new analysis of on-chain transactions reveals the surprisingly isolated nature of NFT phishing attacks and how they exploit complex interactions across multiple blockchain protocols.

Can AI Lie with a Picture? Detecting Deception in Multimodal Models

03.12.2025 by qfx

Subtle deception, even when expressed through combined visual and textual cues, becomes discernible through a system designed to analyze and debate the interplay of these modalities.

Researchers have developed a new method to evaluate whether large language models can convincingly fabricate information when presented with visual cues.

Sharing is Learning: A New Approach to Collaborative AI

03.12.2025 by qfx

A federated learning study demonstrates that KTA v2 achieves 57.7% accuracy on the CIFAR-10 dataset using only approximately 3.8 MB of communication, significantly outperforming FedAvg, which requires roughly 4265.5 MB to reach 42.1% accuracy on the same task.

A novel framework leverages knowledge markets to dramatically reduce communication costs in federated learning, enabling more efficient and effective AI collaboration.

The Research Assistant of the Future: How Close Are We?

03.12.2025 by qfx

Agent performance was evaluated across the DeepResearch Bench (DRB) and a novel system, ourFinder, to assess capabilities in complex research tasks.

A new study dissects the capabilities of AI-powered research agents, revealing key limitations in their ability to synthesize information and conduct genuine inquiry.

Can AI Agents Actually Trade Crypto?

02.12.2025 by qfx

The CryptoBench dataset employs a rigorous, multi-stage verification process for question template creation, followed by a monthly pipeline that generates novel, solvable questions from this established pool-ensuring the benchmark remains both timely and mathematically sound.

New research reveals that while AI agents are adept at gathering cryptocurrency data, they fall short when it comes to the complex reasoning needed for expert-level financial analysis.

Decoding the Human Chess Player

02.12.2025 by qfx

The chess move prediction framework dynamically selects a model calibrated to a player’s skill level, ensuring accurate move anticipation despite the inherent complexity of the game.

New research leverages AI to predict moves not by calculating the best play, but by modeling the behavioral patterns of players at different skill levels.

Beyond the Buzzwords: AI Sentiment Analysis for Finance

02.12.2025 by qfx

The analysis of financial and social sentiment across diverse datasets - including FinancialPhraseBank, Financial Question Answering, Gold News Sentiment, Twitter Sentiment, and Chinese Sentiment - demonstrates a varied distribution of classes within each corpus, suggesting differing characteristics in the expression of sentiment across these platforms and languages.

New research shows that powerful language AI can accurately gauge financial sentiment, even with limited data, challenging traditional domain-specific approaches.

Beyond Textbooks: The Rise of AI-Powered Learning Materials

02.12.2025 by qfx

Podcast script generation, assessed through human evaluation across two academic papers, demonstrates the potential-and inherent limitations-of automated content creation as a means of disseminating complex research.

New research demonstrates the potential of advanced artificial intelligence to create diverse educational content, moving beyond traditional text-based approaches.