Long-Context Processing
The capability defined by the upper limit on the number of tokens a large language model can process in a single inference pass, enabling the model to consult multiple long documents within one inference and reducing the fragmentation of information that results from chunked retrieval.
Timeline (last 90 days)
Early large language models typically had context windows of only a few thousand tokens, requiring complex spreadsheets to be sliced and fed into the model in batches.
A tax workbook with 50 tabs is a stress test of the model's long-context capabilities, challenging both the scale of data it can accommodate and the precision of long-range dependency reasoning.
All Facts (2)
Early large language models typically had context windows of only a few thousand tokens, requiring complex spreadsheets to be sliced and fed into the model in batches.
50%UnverifiedA tax workbook with 50 tabs is a stress test of the model's long-context capabilities, challenging both the scale of data it can accommodate and the precision of long-range dependency reasoning.
50%