Portfolio Optimization

In the volatile world of crypto, the only constant is change. This makes portfolio optimization a challenge — especially for traditional strategies that assume stability over time. A new study by Castelli, Giudici, and Piergallini offers a compelling solution: build your investment pipeline out of agents. Using a modular Multi-Agent System (MAS) framework implemented in Crew AI, the authors compare two crypto portfolio strategies over the 2020–2025 period: one static and one adaptive. The system orchestrates specialized agents to ingest, clean, analyze, optimize, and report on daily crypto prices — all in a transparent and auditable way. ...

In asset management, few debates are more charged than the tug-of-war between causal purity and predictive utility. For years, a growing number of voices in empirical finance have argued that causal factor models are a necessary condition for portfolio efficiency. If a model omits a confounder, the logic goes, directional failure and Sharpe ratio collapse are inevitable. But what if this is more myth than mathematical law? A recent paper titled “The Myth of Causal Necessity” by Alejandro Rodriguez Dominguez delivers a sharp counterpunch to this orthodoxy. Through formal derivations and simulation-based counterexamples, it exposes the fragility of the causal necessity argument and makes the case that predictive models can remain both viable and efficient even when structurally misspecified. ...

In quantitative finance, sparse portfolio optimization is a famously unforgiving problem. Selecting the top m assets from a universe of n under budget and risk constraints is NP-hard, highly sensitive to hyperparameters, and often brittle in volatile markets. Traditional solutions—from greedy algorithms to convex relaxations—either crumble under market shifts or produce opaque, overfitted outputs. But what if we reframed the problem entirely? Enter EFS (Evolutionary Factor Search), a radical new framework that turns sparse portfolio construction into an LLM-guided ranking game. Instead of laboriously tuning machine learning models or relying on rigid heuristics, EFS lets large language models generate, evolve, and select alpha factors—and it does so in a way that is not just automated, but interpretable, adaptive, and surprisingly effective. ...

When markets move, they do so with both sudden shocks and slow drifts. Yet for years, much of optimal trading theory has treated volatility as if it were static—a constant backdrop rather than a dynamic participant in the game. The recent paper by Chan, Sircar, and Zimbidis decisively challenges that assumption by embedding multiscale stochastic volatility into a classical dynamic trading model. The result? A more nuanced, volatility-aware framework that adapts trading speed and target positions based on the fast and slow undulations of risk. ...

The idea of merging language models and financial algorithms isn’t new — but HARLF takes it a step further by embedding them in a hierarchical reinforcement learning (HRL) framework that actually delivers. With a stunning 26% annualized ROI and a Sharpe ratio of 1.2, this isn’t just another LLM-meets-finance paper. It’s a blueprint for how sentiment and structure can be synergistically harnessed. From FinBERT to Fortune: Integrating Text with Tickers Most financial LLM pipelines stop at score generation: classify sentiment and call it a signal. But HARLF builds a full sentiment pipeline using FinBERT, generating monthly sentiment scores from scraped Google News articles for each of 14 assets. These scores aren’t just inputs — they form a complete observation vector that includes: ...

What if you could simulate thousands of realistic futures for the market, all conditioned on what’s happening today—and then train an investment strategy on those futures? That’s the central idea behind a bold new approach to portfolio optimization that blends score-based diffusion models with reinforcement learning, and it’s showing results that beat classic benchmarks like the S&P 500 and traditional Markowitz portfolios. ...

The search for the elusive optimal portfolio has always been a balancing act between signal and noise. Covariance matrices, central to risk estimation, are notoriously fragile in high dimensions. Classical fixes like shrinkage, spectral filtering, or factor models have all offered partial answers. But a new paper by Bongiorno, Manolakis, and Mantegna proposes something different: a rotation-invariant, end-to-end neural network that learns the inverse covariance matrix directly from historical returns — and does so better than the best analytical techniques, even under realistic trading constraints. ...