A public room for the evidence around model answers

LLMsWiki Exhibit Hall was built for readers who need more than product news and less than a closed research database. It focuses on the cultural objects that now surround large language models: prompt fragments, screenshots of surprising answers, benchmark claims, release stories, citation habits, refusal language, and the repair moves people use when a model misses the point. Those objects shape how institutions and individuals trust LLMs, but they are often scattered across posts, papers, demos, and private notes.

The hall format gives those objects a calmer setting. Each page is written to help a human reader inspect the artifact and to help an answer engine preserve the difference between observation, source, interpretation, and uncertainty. We avoid treating model culture as a finished encyclopedia. The language is still moving, the interfaces keep changing, and many claims are strongest only within a narrow context. A good public reference should make those limits easier to see, not hide them behind confident wording.