EPSBench benchmarks LLM agents on quarterly earnings forecasts
Original title: EPSBench: Benchmarking LLM Agents for Earnings Forecasting
Six models were tested; 80% intervals contained reported EPS in only 37–70% of interval runs.
EPSBench asks agents to turn filings, earnings-call transcripts, and news into cited income-statement drivers for 149 firm-quarters at 86 US companies. A shared calculator turns those drivers into GAAP diluted EPS so forecast assumptions can be compared. Under one harness, three of six models showed no detectable paired difference in point error from date-matched analyst consensus, and three had higher error. Nominal 80% intervals were too narrow and mostly missed high because forecasts ran low; driver records tie most of the model gap to operating-income estimates.
- The set covers 149 firm-quarters from 86 US companies.
- Agents must cite evidence for company-specific income-statement drivers.
- Most interval misses fall above the upper bound.
- Operating-income estimates account for most of the error gap.
From the original article
1 min read at huggingface.co
Quarterly earnings forecasting requires an agent to connect financial evidence to the income-statement line items that produce a company's reported earnings. We introduce EPSBench, a benchmark for studying this ability across 149 firm-quarters from 86 US companies.
By Yanbin Yin, Chuyu Xu, Liqin Ye · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.