analiza aiqbench
This study addresses the critical issue of hindsight bias in financial language models by training NoLBERT exclusively on pre-1996 text to evaluate its ability to forecast stock returns without data leakage. The research demonstrates that while combining multiple topic modeling approaches can yield a statistically significant Sharpe ratio against a zero-return benchmark, it fails to outperform simple price-only benchmarks after accounting for search costs. Ultimately, the work highlights the limitations of current NLP techniques in generating robust alpha from news sentiment compared to traditional quantitative methods.
Reading the News Without Hindsight: NoLBERT Topics and the Cross-Section of Stock Returns
arXiv:2610.22745v1 Announce Type: cross Abstract: Language models that read financial news are usually trained on text from the same years used to test them, so a backtest can reward hindsight. We build news topics from NoLBERT, an encoder pretrained only on text published between 1976 and 1995, adapted into a sentence encoder by contrastive fine-tuning on passages of the same article, with no labels or returns. We compare it with LDA, MiniLM and FinBERT readers on 1.26 million articles from 2012 to 2024, convert each reader's topics into attention shocks and stock exposures, and forecast the returns of the 1,000 largest US stocks with eight models at three rebalancing frequencies. Models are fitted on 2013-2019, choices are made on 2020-2021, and 2022-2024 is held out. The experiment with the best validation score loses money at every frequency. Measuring attention shocks against the past 63 trading days instead of 5 raises the test rank IC of every reader. NoLBERT gives the most stable topics and forecasts about as well as LDA. An equal-weight combination of all readers and models earns a monthly net Sharpe ratio of 1.12, significant against zero before adjusting for the search but not against a price-only benchmark.
Abstract:Language models that read financial news are usually trained on text from the same years used to test them, so a backtest can reward hindsight. We build news topics from NoLBERT, an encoder pretrained only on text published between 1976 and 1995, adapted into a sentence encoder by contrastive fine-tuning on passages of the same article, with no labels or returns. We compare it with LDA, MiniLM and FinBERT readers on 1.26 million articles from 2012 to 2024, convert each reader's topics into attention shocks and stock exposures, and forecast the returns of the 1,000 largest US stocks with eight models at three rebalancing frequencies. Models are fitted on 2013-2019, choices are made on 2020-2021, and 2022-2024 is held out. The experiment with the best validation score loses money at every frequency. Measuring attention shocks against the past 63 trading days instead of 5 raises the test rank IC of every reader. NoLBERT gives the most stable topics and forecasts about as well as LDA. An equal-weight combination of all readers and models earns a monthly net Sharpe ratio of 1.12, significant against zero before adjusting for the search but not against a price-only benchmark.
Statistical Finance (q-fin.ST); Computation and Language (cs.CL); Machine Learning (cs.LG); Portfolio Management (q-fin.PM)
arXiv:2610.22745 [q-fin.ST]
(or arXiv:2610.22745v1 [q-fin.ST] for this version)
https://doi.org/10.48550/arXiv.2610.22745 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 299 cuvinte
Subiecte conexe
Why I Left Google DeepMind
LessWrong · 15 iul. 2026
The current bottleneck is political will, not research
AI Alignment Forum · 11 iul. 2026
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face
AI Alignment Forum · 3 aug. 2026
Dialogue with Eliezer Yudkowsky on Neural Networks
LessWrong · 2 oct. 2026