Tech
"You’re right"—LLMs will get climate info wrong if you ask
Publication: Ars Technica
Author: Scott K. Johnson
Source published: 19/11/2026, 18:44:39 UTC
Briefing created: 19/11/2026, 19:45:00 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 19/11/2026, 19:13:23 UTC.
Briefings
A University College London and UNESCO study tested 33 large language models on climate Q&A, revealing that models often fail to provide accurate climate information even when users believe they are correct. The research used benchmark questions and complex interactions to evaluate model outputs, highlighting a significant gap between model capabilities and user expectations regarding climate science accuracy.
Key points
- A study by University College London and UNESCO tested 33 large language models on climate Q&A, revealing that models often fail to provide accurate climate information even when users believe they are correct.
- The research used benchmark questions and complex interactions to evaluate model outputs, highlighting a significant gap between model capabilities and user expectations regarding climate science accuracy.
Context
This research addresses the growing reliance on large language models for public information, examining how they handle climate change queries. The study is part of ongoing efforts to understand the reliability of AI-generated climate information, though it does not specify the exact date of the initial model release or the specific timeframe of the user interactions analyzed.
Limitations
The study's findings are limited to the specific models tested, which may not represent all frontier models released after the experiment concluded. The data does not include information about the exact date of the initial model release or the specific timeframe of the user interactions analyzed, limiting the generalizability of the results.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 19/11/2026, 19:45:00 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeihdpgnobk5mwz3ywc34atuky6pqlqyjst6whx7yadmdwrlbo4pctq