Focused on AI applications in the fintech sector
This paper introduces F²Agent, a multimodal trading system using four specialized agents (market, technical, sentiment, news) with adaptive fusion and noise-robust regularization, achieving over 20% average annualized return improvement across six assets versus 16 baselines.
This paper introduces CoLAS, a framework that explicitly models multimodal corroboration—non-cancelling support across heterogeneous financial modalities—to extract more reliable trading signals, achieving over 25% average annualized return improvement across six datasets including five stocks and Bitcoin.
FinSMART introduces a market-aligned reinforcement learning framework that trains LLMs for financial sentiment analysis using realized market returns instead of human annotations, employing GRPO with a dual-filter trading reward to achieve 220% higher cumulative returns than the previous state-of-the-art FinDPO.
AlphaSchema introduces a structured trading semantic space with five-dimensional schema plans to decouple exploration from LLM implementation, using a surrogate model and quota-based selection to discover high-performing factors that outperform baselines on China A-share markets while demonstrating robustness across LLM backends.
MAPLE introduces a backbone-agnostic framework that generates multiple diverse alpha signals in a single training run by combining a unified capacity-scaled prediction head, extreme rank-weighted listwise loss, and explicit diversity regularization, achieving superior Sharpe and Calmar ratios across four global equity markets with 55x fewer parameters than MoE baselines.
A new arXiv paper introduces Information Fidelity to measure how LLM compression of 10-Q MD&A sections and earnings calls changes downstream investment decisions, identifying decontextualization and model dependency as key failure modes and proposing Agentic Context Compression (ACC) to audit and reduce decision flips.
XAlpha introduces a memory-driven AI quantitative researcher that automates the full hypothesis-to-code alpha discovery loop via a multi-brain architecture, integrating report knowledge, hypothesis planning, factor evolution, empirical validation, and feedback integration, achieving superior performance on CSI300.
EvoQuant introduces a self‑evolving, validator‑guided framework that lets large language models diagnose, propose, and verify improvements to existing quantitative trading strategies, achieving Sharpe ratio gains of up to 199% across A‑share and cryptocurrency benchmarks while demonstrating robustness through ablation and walk‑forward tests.
The article analyzes MetaPS, a simulation‑guided framework that adaptively selects executable trading programs from a strategy library, showing that supervised meta‑strategy learning improves returns across 0.8B‑9B parameter models and outperforms fixed‑strategy baselines, direct decision agents, and prompt‑based LLM agents in both stock and sandbox environments.
AlphaMemo introduces a structured‑search‑process memory for LLM agents that records effective and failed edit patterns in specific parent‑factor contexts, uses AST‑difference extraction, confidence‑gated residual learning, and asymmetric veto to tackle combinatorial search, noisy feedback, redundancy, and over‑fitting, achieving superior out‑of‑sample performance and discovery efficiency on CSI 500 and S&P 500 benchmarks.
