Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Needle: The benchmark your search engine can't memorize

keenable.ai

Post reply on HN
Print view View original

11–14 of 14 posts

Re: Needle: The benchmark your search engine can't memorize

#11

Post by matt4711 »

mpalmer wrote: post #9

The blog post appears to get confused and devotes its entire second half to pitching Keenable itself. If the idea is to build credibility for the new benchmark, this maybe was not the best choice. Search engines with their own indexes can retrieve documents no one else has. That means your agents will be able to answer questions previously unanswerable. Besides the clear AI smell, this nonsensical claim also plainly…

> Besides the clear AI smell, this nonsensical claim also plainly contradicts the methodology's key evaluation claim that the quality of an engine's results should be measured against how much it overlaps with the reranked aggregate of the other engines. The benchmark thus seemingly values an engine's ability to "answer unanswerable questions" at zero.

The engine itself is part of the reranked aggregate so if it finds something useful and everybody else does not it gets full credit.

Re: Needle: The benchmark your search engine can't memorize

#12

Post by daft_pink »

so is this an independent search engine benchmark or a blog post from a search engine provider showing their search engine at the top of the benchmark?

i'm a little bit confused as at first when I was reading it i thought it was a search engine benchmark, but it seems that keenable is at the top which i assume is related to the web domain owner? i've never heard of keeenable

Re: Needle: The benchmark your search engine can't memorize

#14

Post by phoenix120 »

mpalmer wrote: post #9

The blog post appears to get confused and devotes its entire second half to pitching Keenable itself. If the idea is to build credibility for the new benchmark, this maybe was not the best choice. Search engines with their own indexes can retrieve documents no one else has. That means your agents will be able to answer questions previously unanswerable. Besides the clear AI smell, this nonsensical claim also plainly…

> The blog post appears to get confused and devotes its entire second half to pitching Keenable itself. If the idea is to build credibility for the new benchmark, this maybe was not the best choice.

The exact sentence you cite is an example that I've tried everything I could to avoid that. In one of the original revisions it was "We can retrieve very unique documents no one else has", not "Search engines with their own indexes can retrieve documents no one else has".

> Besides the clear AI smell, this nonsensical claim also plainly contradicts the methodology's key evaluation claim that the quality of an engine's results should be measured against how much it overlaps with the reranked aggregate of the other engines. The benchmark thus seemingly values an engine's ability to "answer unanswerable questions" at zero.

It is fun to hear "AI smell" accusations about these specific sentences. I do use Claude to fix my grammar mistakes and give editorial opinions, mostly because I'm not a native English speaker. And as for these specific sentences, I even had a session where Claude fixed an obvious mistake in my writing: https://snipboard.io/rkm4LD.jpg

There is no contradiction to the methodology. We use ultimate engine to calculate IDCG, but that doesn't mean that the methodology favors the overlap. 5 relevant unique results would produce exactly the same nDCG as 5 relevant non-unique results.

The main part of the benchmark (the one that measures quality) doesn't care about result uniqueness (if it is not about full URL duplicates).

> Yeah? Care to cite anything for that?

The closest source is probably LRAT (https://arxiv.org/abs/2604.04949), it is cited in the benchmark, we used their trajectories in AgenticRare.

Post reply on HN