To make it easy to resolve these I'll give both ChatGPT and Perplexity the market description and ask how the answers should resolve. If they disagree I'll ask Gemini. If the llms refuse to make a decisive choice I'll resolve the answer N/A.
I'll use whatever the standard model is that comes without a paid plan and has web access. I'll resolve no sooner than the 15th so that adequate data can appear online.
Update 2025-06-14 (PST) (AI summary of creator comment): The creator will post the LLM resolutions in a comment before resolving the market.
The market will remain open for at least one day after this to allow users to provide evidence of potential LLM hallucinations.
If an answer is controversial, the creator will still resolve according to the LLM output as described in the original criteria.
Update 2025-06-14 (PST) (AI summary of creator comment): Events must happen geographically at the parade in DC. An event's causal link to the parade is not sufficient if it is not geographically present (e.g., an event in LA would not count).
Update 2025-06-17 (PST) (AI summary of creator comment): The creator will resolve the 'Trump compliments himself' answer as NO, overriding the established LLM resolution process for this specific case.
The creator is using their own judgment, stating that the LLMs' reasoning for a YES resolution was unreasonable, despite both Perplexity and the Gemini tiebreaker agreeing on YES.
Update 2025-06-17 (PST) (AI summary of creator comment): In response to a user question, the creator confirmed they will correct an LLM's judgment if they believe it is wrong. This establishes that the creator may use their own judgment to override the established LLM resolution process.
Update 2025-06-18 (PST) (AI summary of creator comment): In response to a user question, the creator has stated that an event occurring two days after the parade is considered too late to be included in the resolution.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ597 | |
| 2 | Ṁ581 | |
| 3 | Ṁ264 | |
| 4 | Ṁ211 | |
| 5 | Ṁ209 |
People are also trading
@traders At this point I will resolve all answers as NO. I'll keep this open for at least 24 more hours for anyone to make a case for different resolution.
The ChatGPT and Perplexity agree on all but
- Trump compliments himself (ChatGTP: NO, Perplexity: YES)
- At least 50 protestors are arrested (ChatGTP: YES, Perplexity: NO)
Gemini pulled in as a tiebreaker on these two items agree with Perplexity on NO for the 50 protestor arrests and I do as well. ChatGPT based its YES on a article about 60 arrests from the day before the parade.
Gemini agrees with with Perplexity on Trump complimenting himself. Both Gemini's and Perplexity's reasoning is based on the idea that holding the parade at all is a self-compliment. However, given the context of this market, I think this is unreasonable and would like to resolve this NO regardless. We knew the parade would take place when I created this market and given Trump's baseline of self-complimenting this barely registers.
## LLM Chats:
ChatGPT: https://chatgpt.com/share/6851e790-bdf8-8004-91ee-14e94d91b2d0
Perplexity: https://www.perplexity.ai/search/for-a-prediction-market-with-t-WO2rV85PSB2P418yi_ob1Q
Gemini as PDF based on GDocs export of its output (I couldn't find a way to share the whole chat as with the other LLMs): https://personel.s3.eu-west-1.amazonaws.com/Resolving+Trump+Parade+Outcomes+-+Google+Docs.pdf
https://www.thedailybeast.com/truck-carrying-army-tank-for-trumps-parade-kills-pedestrian-in-dc/
Someone dies should be YES
Thanks for taking to into account!
@Marnix Am I understanding correctly that the number of protestors physically present at the parade must be estimated as 100,000 or more (by a credible source, presumably)?
(im taking back my statement I misread the question)
@AlexanderTheGreater yeah but it’s in response to the parade it’s happening simultaneously and it wouldn’t have happened if the parade wasn’t happening In my eyes that’s causal
@Ebcc1 as per title, it's gotta be at the parade. I'm honestly worried about ambiguity of how many blocks in DC count as "at the parade", but I think it's safe to say that there are too many blocks between the parade and LA.
@AlexanderTheGreater can you please not only rely on LLMs for the resolution? We all know their chance of hallucinating or misusing data
@tobiasscheuer when asking about recent events the biggest problem is sensationalist headlines the are then used by LLM search.
@tobiasscheuer so many recent markets that are related to Trump have turned into markets that are about the judgement of the market creator. I absolutely don't want to be in that position. That's why I went for having a llm resolve it. Otherwise I wouldn't have created the market and in fact originally decided against creating it, till I had the llm idea.
That said, how about this: I'll post the llm resolutions in a comment and keep the market open for at least a day day after that to give opportunity to call it hallucinations and provide evidence. However, if any answer is controversial I'll blindly go with the resolution via llm as described in the original criteria.
Does that sound like a fair compromise?
@AlexanderTheGreater that sounds fair. Although I would suggest n/a those resolutions where LLMs and evidence are lacking clarity. LLMs can only work with existing texts, they are not truth machines
@AlexanderTheGreater give it as much time as it takes. Don't feel pressured to resolve immediately if the answer is not clear yet.