Predicting Stock Price Direction on Earnings Announcement Days using Multi-modal Deep Learning
Researchers demonstrate that combining FinBERT-derived sentiment signals with fundamental and technical market data improves directional stock price forecasting on earnings announcement days. The study benchmarks LSTM and Transformer architectures against logistic regression, isolating sentiment's incremental predictive power in a high-noise financial domain. This work exemplifies how domain-specific language models and multi-modal fusion are reshaping quantitative finance, though real-world deployment challenges around data leakage and market microstructure remain unaddressed.
Modelwire context
Skeptical readThe paper isolates sentiment's contribution via controlled benchmarking, but doesn't address whether FinBERT's earnings-day signal survives transaction costs, execution slippage, or the data leakage risk that arises when sentiment and price move together intraday. The 'incremental predictive power' claim needs a denominator: how much of the gain evaporates once you account for realistic trading friction?
This connects directly to the STaT paper from the same day, which flagged a parallel problem in multimodal forecasting: models that minimize average error often produce overly smooth predictions that miss the turning points practitioners actually need. Here, the researchers are adding a sentiment layer to capture directional moves on high-volatility announcement days, but they haven't validated whether their model preserves the sharp transitions that make the prediction actionable. The essay scoring work from today also surfaces a related risk: domain-adaptive pretraining on learner corpora showed mixed results when applied to downstream tests, suggesting that fine-tuning FinBERT on financial text may not transfer cleanly to the specific microstructure of earnings announcements.
If the authors release backtests on a held-out earnings calendar from 2025 (not in their training window) and show the sentiment signal persists after controlling for pre-announcement volatility spikes and post-market reversals, the result moves from academic to credible. If they don't, or if the gain shrinks below 2-3% directional accuracy improvement, the work remains a proof-of-concept rather than a deployment candidate.
Coverage we drew on
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsFinBERT · LSTM · Transformer · arXiv
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.