Combining Constrained and Unconstrained Decoding via Boosting: BoostCD and Its Application to Information Extraction
Marija Šakota, Robert West · 2025
Read Combining Constrained and Unconstrained Decoding via Boosting: BoostCD and Its Application to Information Extraction by Marija Šakota, Robert West (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-14901.txt.
Description
Marija Šakota and Robert West present BoostCD in a 2025 arXiv paper. Language models often must output structured objects, and constrained decoding forces the output to fit the structure. Yet models trained without the constraints can falter when forced at test time.
BoostCD decodes twice in a first phase, once with constraints and once without. A second, learned model then merges the two drafts into one final answer. The errors of the two modes tend to differ, and the merger learns to exploit that. Applied to closed information extraction, their model beat prior approaches in and out of distribution.
Catalog record
| CREATOR | Marija Šakota, Robert West |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-14901