LLMs beat humans at cross-modal associations: Gemini ~50% vs human ~23% accuracy
abenitezburraco · x · 2026-11-15
A study shared on X reports that Large Language Models outperform humans in cross-modal association tasks. The accompanying chart shows accuracy of roughly 23% for humans, 41% for ChatGPT, and 50% for Gemini, with both models clearly ahead of human raters and Gemini on top — suggesting current LLMs may already surpass humans at linking concepts across modalities.
More from Models
- Why do LLMs always end replies with a question? Reddit users suspect engagement tuning — TaurusGold · 2026-11-15
- Chart shows Gemini 2.5 Pro sliding from LMArena #1 to #42 in about a year — Sauers_ · 2026-11-15
- Dev says Google Antigravity's Hydrogen model is "VERY good and VERY cost effective" — jasonkneen · 2026-11-15
- Grok 4.8 reportedly rolling out to Grok Bot users, unconfirmed yet — mark_k · 2026-11-15
- Llama 405B Base Model Natively Mimics o1's Chain-of-Thought from 128 Pages of Context — Sauers_ · 2026-11-15
- Gemini 4 'Hydrogen' Rumored Faster Than GPT and Claude, a Third Model Class — mertdumenci · 2026-11-15