Tip

Paper Reveals JSON/XML Formats Reduce Model Response Diversity

Published: 2026-07-27 23:38 Source: X: Vista (@vista8)

A paper testing 44 models found that requiring output in JSON or XML format significantly reduces the diversity of large language model responses. Models are reinforced with large amounts of structured data during post-training, naturally tending to give "standard answers," thereby suppressing output richness.

Read the original (opens in a new tab)

News stream data aggregated by AI HOT

Related newsLatest in this category

Tip Slopcodebench to Become New Threshold for AI Conversations

07-28 10:11 · X: Dex Horthy (HumanLayer) (@dexhorthy)

Tip Knowing the names of things in the AI era is like a superpower

07-28 10:06 · X: Ethan Mollick (@emollick)

Tip Anthropic CEO Responds to NVIDIA Joint Open Letter, Advocates Alternative Measures Instead of Banning Open-Weight Models

07-28 09:51 · X: Shao Meng (@shao__meng)

Tip Terence Tao Talks AI at International Congress of Mathematicians: AI Proofs Will Increase, but Math Won't Get Faster https://best.xiaohu.ai/article/tao-age-of-ai-icm-2026/

07-28 09:38 · X: Xiaohu (@xiaohu)

Tip Opus 5 Benchmark Results on SlopCodeBench

07-28 09:29 · Hacker News Hot (buzzing.cc Chinese Translation)