Breaking01 / AI
TETHER maps long documents from their tables of contents
TETHER matches text blocks to table of contents entries to find where chapters and sections begin and end, without additional training.

Yu Zou and colleagues reported in a paper submitted to the research repository arXiv on 11 Nov 2026 that TETHER used 90% fewer tokens, or units of text, than methods that extract structure directly from long documents.
TETHER uses a model that reads text and images to work out how chapters and sections fit together from the table of contents. It then links blocks of text to those entries to find where each chapter and section begins and ends. The method requires no additional training.
The researchers evaluated TETHER on documents in several languages, including product manuals and annual reports. In their published benchmark, TETHER recovered document structure more accurately than existing document parsers.
The paper’s abstract did not disclose token savings by document type.