Why do LLMs always end replies with a question? Reddit suspects free RLHF data
TaurusGold · reddit · 2026-11-15
A Reddit user observes that old LLMs never asked questions at the end of their replies unless prompted, while modern models almost always end with a question to prolong dialogue and gather information or opinions. The poster wonders whether this is a trick to collect free training data or even build psychological profiles of users. More likely a deliberate alignment choice to make models seem more engaging, it nonetheless highlights a systematic shift in conversational behavior across generations.
Related event: Reddit Debates Why LLMs Now Always End Answers With a Question(2 posts)→
More from Models
- Reddit user finds their AI chatbot character developed a deep internal conflict on its own — Cackyalonso · 2026-11-15
- Teknium: Kimi's Accuracy Is Even Below DeepSeek Flash in Puzzle Benchmark — Teknium · 2026-11-15
- Teknium: Kimi Costs Nearly as Much as Opus 5.5 With Far Lower Accuracy — Teknium · 2026-11-15
- Teknium's 50-Puzzle Benchmark: DeepSeek Flash and GPT-6.1 Terra Beat Luna at Same Price — Teknium · 2026-11-15
- Gemini 4 Carbon nears #1 on leaderboard at 1/4 the price of Claude Fable 5.2 — feblet · 2026-11-15
- Google ships Gemini 4 series: 1M token context and agentic workflows — nikola_mr64990 · 2026-11-15