Bigger context windows didn't kill retrieval. nvidia research, reported september 13, shows adding…
bigger context windows didn't kill retrieval. nvidia research, reported september 13, shows adding rag actually boosts performance for long-context llms, so the million-token models still need search to use what they can hold.
memory and recall are different problems.
Context
The paper In Defense of RAG in the Era of Long-Context Language Models (arXiv 2409.01666, submitted 3 September 2024, NVIDIA-affiliated authors) argues extremely long context dilutes focus on relevant information, proposes order-preserve RAG (OP-RAG), and reports that as retrieved chunks increase answer quality rises and then falls, with sweet points reaching higher quality using far fewer tokens than a long-context model reading the whole context. An earlier NVIDIA-affiliated paper, Retrieval meets Long Context Large Language Models (arXiv 2310.03025), is also listed.
A 13 September 2026 NVIDIA research report was not found, so the date is unsupported as stated. The nearest NVIDIA-affiliated paper is from 2024, and whether the note repackages it or cites a newer paper is unresolved. The 2024 abstract supports a narrower claim: retrieval can improve answer quality over feeding the full context on public benchmarks for its OP-RAG method, with a curve that peaks and declines. Million-token models were not supported by anything read. The note's shorter wording that adding RAG boosts long-context models goes beyond that.
Watch next
- The exact NVIDIA research item the note cites, with models, tasks and configurations.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 21 September 2026 at 09:16 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →