6,000 fruit picks: ChatGPT's "random" choice changes sharply by language
Raffallos · reddit · 2026-08-24
A developer asked GPT-5.6-Luna to randomly pick a fruit from a 10-item list 6,000 times each in English, Polish, and Japanese. Despite identical wording and fruit order, only four fruits were ever picked across all three languages.
A second run of 6,000 trials with the fruit list shuffled each time showed slightly better—but still only three—fruits selected. Full write-up, GitHub methodology, and the raw dataset on Hugging Face are linked.
More from Models
- Users frustrated as Gemini web app access doesn't carry over to coding tools — dreamwieber · 2026-10-16
- OpenAI simplifies API usage tiers into Build/Launch/Grow with automatic upgrades at $5/$100/$500 — prd_008 · 2026-10-16
- Xiaomi's MiMo RL dashboard reveals $3.47M spent on just 30 training steps — bycloud · 2026-10-16
- Analyst: in a year, models 2-3x better across the board at 10x lower cost, token use up 10-20x — DeryaTR_ · 2026-10-16
- Google revamps Gemma docs, site teases Gemma 4 open models — rseroter · 2026-10-16
- Respan's Span-02 brings anytime reasoning: interrupt it mid-thought and still get the best answer — techNmak · 2026-10-16