DistillSem: LLM-Distilled Financial Headline Semantics for Cross-Sectional Return Prediction
“Using 96,802 teacher labels spanning 2015 to 2022, our best distilled representation yields an information coefficient (IC) of 0.0225 (t = 3.80) on a heldout cross-sectional return ranking task covering 2024 to 2026.”
AssetEquities
StrategyMachine learningnews sentiment
TechniqueLarge language model
FrequencyDaily
Also stated: 2015–2022, Text and news
Missing from abstract: benchmark, out-of-sample test, costs, multiple testing, universe
Institutional-grade audit of this strategy.
£5 — Request the audit- Audit of the paper claims accounting for costs, luck, crises, originality, geography, etc.
- Reproduction code integratable into Bloomberg, IBKR, etc.
- Bias-free datasets for backtesting included
- Delivered in 2 working days
The paper's neighbourhoodMachine learning return prediction by year
The family is counted across the whole universe. Asset, method and authors are counted inside the family.
| Similar papers | citation | same method | same asset | same authors | |
|---|---|---|---|---|---|
| Deep learning-based analysis of investor sentiment and stock return prediction: an experimental finance perspective by Zheng et al. | 2026 | ||||
| Autonomous Market Intelligence: Agentic AI Nowcasting Predicts Stock Returns by Chen and Pu | 2026 | ||||
| Cross-Stock Predictability via LLM-Augmented Semantic Networks by Huang et al. | 2026 | ||||
| Structured Event Representation and Stock Return Predictability by Li et al. | 2025 | ||||
| When Reasoning Fails: Evaluating 'Thinking' LLMs for Stock Prediction by Sodha | 2025 |