parthknowsai
Region: US
Friday 18 September 2026 12:22:00 GMT
23139
1481
53
190
Music
Download
Comments
Lameees :
Queens invented table lookup
2026-09-18 18:19:38
0
kab :
could you add the link to this article please
2026-09-18 13:13:47
2
Romy :
it also has another limitation sometimes a query's response can be in multiple sections not just one
2026-09-18 13:04:38
25
Joshua :
This was already a thing with PageView repo. and it's not RAG
2026-09-18 13:55:20
13
hassanjaved1663 :
This should augment the conventional Rag otherwise you’d miss something critical on the footnote on page 269
2026-09-18 18:24:58
0
Ric Moore :
Jev doesn’t hallucinate
2026-09-18 13:35:28
1
Olusola :
So the solution have been in our faces
2026-09-18 18:17:40
0
Wethotshame :
No vector db?
2026-09-18 18:20:20
0
maplebrown :
RAG in big 2026 😹 just use the 1 gazillion token context window bros
2026-09-18 16:24:50
2
Galaxy_eye :
I came up with this 3 months ago. 100% accuracy.
2026-09-18 17:17:09
1
OptiPrime AI :
I am building in the law space. My hybrid RAG design has no hallucinations and a 85% average accuracy. It uses re-ranking which improves accuracy, reduces hallucinations, but it is slow, about 6 minutes, but it can work across messy docs like 15 year old court docs.
2026-09-18 14:09:05
0
Capel :
Wouldn’t be a limitation if the document gets updated? It can generalize and keep retrieval quality stable until certain point?
2026-09-18 14:20:38
0
🪨 :
Do you have a video on lev yet?
2026-09-18 12:32:22
0
randomizer69578 :
So, an indexed database 😅
2026-09-18 15:12:48
1
qwertz_enjoyer :
assuming there is a toc
2026-09-18 14:26:16
0
idite naxui :
omg that's just what I needed, thank you!
2026-09-18 12:38:44
0
Le chaton :
So how is that different from having docs OCR into markdown and then have any llm with the proper prompt go read ToC then the section ?
2026-09-18 15:23:46
1
Cris :
table of contents isn't basically what Traditional database do when doing a Index for a DB but either way how does it check that the title for the embedded content isn't misleading?
2026-09-18 14:03:47
2
NS 🕷️ :
Vectorless rag or page index rag. Not new
2026-09-18 18:36:02
0
viann205 :
what's the name of the paper
2026-09-18 16:54:01
0
Dan.Dusk :
the more i read abot rag alternatives the final conclussion is always "our data is too disorganized" or " we dont want toninvest in cleaning the data"
2026-09-18 18:35:48
0
kevnotfound :
i think i invent or i just make my own concept better than this
2026-09-18 18:08:06
0
Stephen :
I struggle where the LLM isn’t provided accurate chronological data. It’s able to answer pointed questions, but not able to discern progression through time and changes. When the documents include progress, updates, changes, and overall transformation from initial state to final state, it will only pull the valid records based on the query and attempt to answer a question. It’s the difference between reading and referencing a dictionary, vs. interpreting the meaning and context of a historical epic. If that data isn’t in the training data that is.
2026-09-18 17:22:31
0
ElMercader :
Hallucinated “Artificial Table of Contents,” lol 😂
2026-09-18 18:07:58
0
Creatieve Revolte :
Similar to GBrain?
2026-09-18 12:29:38
0
To see more videos from user @parthknowsai, please go to the Tikwm
homepage.
