Bayesian Linguistic Forecaster agent achieves state-of-the-art on forecasting benchmark

Researchers have developed the Bayesian Linguistic Forecaster (BLF), an agentic system designed for binary forecasting tasks. The BLF integrates numerical probability estimates with natural-language evidence summaries, updated iteratively by a large language model. This novel approach has demonstrated state-of-the-art performance on the ForecastBench benchmark, outperforming existing methods like GPT-5 and Grok-4.20. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel agentic forecasting system that sets a new SOTA on the ForecastBench benchmark, potentially improving prediction accuracy in various domains.

RANK_REASON This is a research paper detailing a new agentic forecasting system. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Bayesian Linguistic Forecaster
• ForecastBench
• GPT-5
• Grok-4.20
• Kevin Murphy

• paper
• other