ChatGPT stops citing Reddit, but training data bias remains
mark_k · x · 2026-08-19
A user observes that ChatGPT has stopped citing Reddit in its responses, noting this is a good move by OpenAI. However, they point out the bigger issue remains: ChatGPT is still trained on Reddit data, meaning Reddit's significant bias continues to influence the model's outputs.
Related event: ChatGPT's Reddit Citations Plummet as Search Mechanism Shifts(13 posts)→
More from Models
- Transluce vets: a year of neuron autointerp work still can't explain complex behaviors — aryaman2020 · 2026-10-16
- Grok Bot Ran a 12-Hour Data Analysis Nonstop, Handling API Rate Limits Itself — Daniel_Farinax · 2026-10-16
- One year of Recursive Language Models: GPT-5-mini beats GPT-5 by 110% on 132k-token tasks, cheaper too — a1zhang · 2026-10-16
- Users Report Claude Models Suddenly More Compliant After Forced Compaction — RealSharpNinja · 2026-10-16
- Prediction: Opus 5.5-level models on laptops by 2027, backed by Qwen 27B data — cephaloform · 2026-10-16
- Users Say Krea 2 Output Degraded Sharply Since Krea 3 Beta Launched — szywis · 2026-10-16