Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by kumama on Wednesday August 05, 2026 @09:56PM
(#49191496)
castform founder here. we should have made it more prominent on the blog but here's the full code example: https://github.com/castform-ai/benchmax/tree/main/examples/n...
Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by sreekanth850 on Wednesday August 05, 2026 @11:52PM
(#49192220)
Are you using blind chunking or section aware chunking?
Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by kumama on Thursday August 06, 2026 @12:10AM
(#49192323)
for the example here the chunking is section aware -> but the general training data synthesis pipeline is agnostic to type of chunking
Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by sreekanth850 on Thursday August 06, 2026 @12:18AM
(#49192375)
if chunking is section aware, how do you manage large section embeddings?
Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by Foobar8568 on Thursday August 06, 2026 @12:24AM
(#49192410)
Thank you for the github link, what is the license? I went briefly over several repo, unless I missed it somewhere the license was often missing.
Re:Beating GPT-5.6 Sol on retrieval with 100x cheaper open m...
by kumama on Thursday August 06, 2026 @12:32AM
(#49192468)
it's apache 2.0: https://github.com/castform-ai/benchmax/tree/main?tab=Apache...
Everything is vibe sloped to death, and dead after a few months to a couple of years (and not hard to be 100 cheaper than GPT-5.6 sol ... DS is basically free and I guess already 100 times cheaper or more, and here another slope ).