EconCausal: A Context-Aware Economic Reasoning Benchmark for Large Language Models

Name: EconCausal: A Context-Aware Economic Reasoning Benchmark for Large Language Models
Uploaded: 2026-10-17T20:18:07.139487+00:00
Duration: 694 s
Description: EconCausal introduces a large-scale benchmark testing whether large language models can correctly infer and revise causal effect directions when the socio-economic context changes. Built from 10,490 triplets from economic studies, it shows that current LLMs struggle with context-dependent reasoning, often failing when signs flip or when given misleading evidence.

EconCausal introduces a large-scale benchmark testing whether large language models can correctly infer and revise causal effect directions when the socio-economic context changes. Built from 10,490 triplets from economic studies, it shows that current LLMs struggle with context-dependent reasoning, often failing when signs flip or when given misleading evidence.

Open the full episode