AutoIndex Learns Executable Representation Programs For Retrieval
Reactions from ranked influencers
6 posts
๐งฉ Every search system has a program that maps: raw documents โ searchable representations That program decides what becomes a retrieval unit, what context is preserved, which signals are amplified, and which document structure the retriever can see. We call this a ๐ฅ๐ฒ๐ฝ๐ฟ๐ฒ๐๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป ๐ฃ๐ฟ๐ผ๐ด๐ฟ๐ฎ๐บ.
I was really surprised how well this worked out of the box, and we're only getting started. So much untapped potential! All of the impressive results and architecture designs should be credited to the UMass MS (!) students here.
๐ Today weโre introducing ๐๐๐๐ผ๐๐ป๐ฑ๐ฒ๐ : Learning Representation Programs for Retrieval. ๐ AutoIndex learns executable programs that determine how a corpus is represented to a search engine. We think this points to a broader category of research: ๐๐ ๐๐๐๐๐ฒ๐บ๐ ๐๐ต๐ฎ๐ ๐ผ๐ฝ๐๐ถ๐บ๐ถ๐๐ฒ ๐๐ต๐ฒ ๐ฒ๐ ๐ฒ๐ฐ๐๐๐ฎ๐ฏ๐น๐ฒ ๐ฝ๐ฟ๐ผ๐ด๐ฟ๐ฎ๐บ๐ ๐ฎ๐ฟ๐ผ๐๐ป๐ฑ ๐๐ต๐ฒ๐บ.
Any inaccuracies can be credit to me. :)
Was a pleasure to work with @mrdrozdov on this, check it out! ๐ Site: https://auto-index.github.io arXiv: https://arxiv.org/abs/2607.18603 code: https://github.com/auto-index/autoindex
๐ Today weโre introducing ๐๐๐๐ผ๐๐ป๐ฑ๐ฒ๐ : Learning Representation Programs for Retrieval. ๐ AutoIndex learns executable programs that determine how a corpus is represented to a search engine. We think this points to a broader category of research: ๐๐ ๐๐๐๐๐ฒ๐บ๐ ๐๐ต๐ฎ๐ ๐ผ๐ฝ๐๐ถ๐บ๐ถ๐๐ฒ ๐๐ต๐ฒ ๐ฒ๐ ๐ฒ๐ฐ๐๐๐ฎ๐ฏ๐น๐ฒ ๐ฝ๐ฟ๐ผ๐ด๐ฟ๐ฎ๐บ๐ ๐ฎ๐ฟ๐ผ๐๐ป๐ฑ ๐๐ต๐ฒ๐บ.
๐ Today weโre introducing ๐๐๐๐ผ๐๐ป๐ฑ๐ฒ๐ : Learning Representation Programs for Retrieval. ๐ AutoIndex learns executable programs that determine how a corpus is represented to a search engine. We think this points to a broader category of research: ๐๐ ๐๐๐๐๐ฒ๐บ๐ ๐๐ต๐ฎ๐ ๐ผ๐ฝ๐๐ถ๐บ๐ถ๐๐ฒ ๐๐ต๐ฒ ๐ฒ๐ ๐ฒ๐ฐ๐๐๐ฎ๐ฏ๐น๐ฒ ๐ฝ๐ฟ๐ผ๐ด๐ฟ๐ฎ๐บ๐ ๐ฎ๐ฟ๐ผ๐๐ป๐ฑ ๐๐ต๐ฒ๐บ.
Joint work with @Sam25491761, @nithyaraj1506, Shreyas Chaudhari, Ramya Narayanasamy, and Hanna Jiang. ๐ Site: https://auto-index.github.io/ ๐ arXiv: https://arxiv.org/abs/2607.18603 ๐ป Code: https://github.com/auto-index/autoindex