feat(web): ask examples that ChatGPT verifiably gets wrong#20
Merged
Conversation
Replace two generic ask examples with questions whose answers live in table/infobox cells — verified empirically: - 'How many shots on target did Inter have in the 2010 Champions League final?' -> 7. GPT-5-class Codex answers 4 WITHOUT web access and STILL answers 4 WITH web search enabled (while citing UEFA/Wikipedia) — text scraping fumbles the match-statistics table. Our agent reads the table tile and answers 7. - 'Which district in Nagaland has the RTO code NL-03?' -> Tuensang. Codex (no web) says Kohima; our agent triangulates NL-01/02/03 across district infoboxes and answers correctly. Both verified end-to-end against the live agent. Kept one accessible classic and the Chinese example for approachability.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Tested 23 SimpleQA-style candidates against Codex (GPT-5-class) as the 'ChatGPT control group'. Modern models have memorized nearly all public-benchmark prose facts — the survivors are table/infobox cells:
The shots question is the flagship: wrong even with browsing, because text extraction mangles the multi-column match-stats table — exactly the failure mode the paper's pipeline figure illustrates. Kept 'Explain The Starry Night' + '介绍一下兵马俑' for approachability.