TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Hard problems that reduce to document ranking

318 pointsby noperator3 months ago

15 comments

obblekk3 months ago
The open source ranking library is really interesting. It&#x27;s using a type of merge sort where the comparator function is an llm comparing (but doing batches &gt;2 for fewer calls).<p>Reducing problems to document ranking is effectively a type of test-time search - also very interesting!<p>I wonder if this approach could be combined with GRPO to create more efficient chain of thought search...<p><a href="https:&#x2F;&#x2F;github.com&#x2F;BishopFox&#x2F;raink?tab=readme-ov-file#description">https:&#x2F;&#x2F;github.com&#x2F;BishopFox&#x2F;raink?tab=readme-ov-file#descri...</a>
评论 #43175701 未加载
antirez3 months ago
One interesting thing about LLMs, that is also related to why chain of thoughts work so well, is that they are good at sampling (saying a lot of things about a problem), and are good, when shown N solutions, to point at the potentially better one. They do these things better than zero-shot &quot;tell me how to do that&quot;. So CoT is searching inside the space of representation + ranking, basically. So this idea is leveraging something LLMs are able to clearly do pretty well.
hexator3 months ago
This furthers an idea I&#x27;ve had recently that we (and the media) are focusing too much on creating value by making more ever more complex LLMs, and instead we are vastly underestimating creative applications of current generation AI.
评论 #43176712 未加载
评论 #43183608 未加载
评论 #43185645 未加载
noperator3 months ago
A concept that I&#x27;ve been thinking about a lot lately: transforming complex problems into document ranking problems to make them easier to solve. LLMs can assist greatly here, as I demonstrated at inaugural DistrictCon this past weekend.
评论 #43175927 未加载
rfurmani3 months ago
Very cool! This is also one of my beliefs in building tools for research, that if you can solve the problem of predicting and ranking the top references for a given idea, then you&#x27;ve learned to understand a lot about problem solving and decomposing problems into their ingredients. I&#x27;ve been pleasantly surprised by how well LLMs can rank relevance, compared to supervised training of a relevancy score. I&#x27;ll read the linked paper (shameless plug, here it is on my research tools site: <a href="https:&#x2F;&#x2F;sugaku.net&#x2F;oa&#x2F;W4401043313&#x2F;" rel="nofollow">https:&#x2F;&#x2F;sugaku.net&#x2F;oa&#x2F;W4401043313&#x2F;</a>)
mskar3 months ago
Great article, I’ve had similar findings! LLM based “document-chunk” ranking is a core feature of PaperQA2 (<a href="https:&#x2F;&#x2F;github.com&#x2F;Future-House&#x2F;paper-qa">https:&#x2F;&#x2F;github.com&#x2F;Future-House&#x2F;paper-qa</a>) and part of why it works so well for scientific Q&amp;A compared to traditional embedding-ranking based RAG systems.
评论 #43176580 未加载
tbrownaw3 months ago
So instead of testing each patch, it&#x27;s faster to &quot;read&quot; it and see if it looks like the right kind of change to be fixing a particular bug. Neat.
adamkhakhar3 months ago
I&#x27;m curious - why is LLM ranking preferred over cosine similarity from an embedding model (in the context of this specific problem)?
评论 #43177127 未加载
patapong3 months ago
Interesting insight, and funny in a way since LLMs themselves can be seen as a specific form of document ranking, i.e. ranking a list of tokens by appropriateness as continuation of a text sequence.
westurner3 months ago
Ranking (information retrieval) <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Ranking_(information_retrieval)" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Ranking_(information_retrieval...</a><p>awesome-generative-information-retrieval &gt; Re-ranking: <a href="https:&#x2F;&#x2F;github.com&#x2F;gabriben&#x2F;awesome-generative-information-retrieval#re-ranking">https:&#x2F;&#x2F;github.com&#x2F;gabriben&#x2F;awesome-generative-information-r...</a>
Everdred2dx3 months ago
Very interesting application of LLMs. Thanks for sharing!
jasonjmcghee3 months ago
I see in the readme you investigated tournament style, but didn&#x27;t see results.<p>How&#x27;d it perform compared to listwise?<p>Also curious about whether you tried schema-based querying to the llm (function calling &#x2F; structured output). I recently tried to have a discussion about this exact topic with someone who posted about pairwise ranking with llms.<p><a href="https:&#x2F;&#x2F;lobste.rs&#x2F;s&#x2F;yxlisx&#x2F;llm_sort_sort_input_lines_semantically#c_xk5zgz" rel="nofollow">https:&#x2F;&#x2F;lobste.rs&#x2F;s&#x2F;yxlisx&#x2F;llm_sort_sort_input_lines_semanti...</a>
marcosdumay3 months ago
Hum... The gotcha is that LLMs can rank for subject relevance, but not for most other kinds of quality.
评论 #43179302 未加载
m3kw93 months ago
That title hurts my head to read
moralestapia3 months ago
Minor nitpick,<p>Should be &quot;document ranking reduces to these hard problems&quot;,<p>I never knew why the convention was like that, it seems backwards to me as well, but that&#x27;s how it is.
评论 #43177909 未加载
评论 #43182438 未加载