LLMs hit an effective binding limit in text-to-optimization as data complexity grows, but externalizing numeric data via BIND and binding-focused finetuning raises accuracy from 59% to 82%+ and lets small specialists match larger end-to-end models.
The structured representation is then passed to an LLM (GPT-5) with a prompt, and all numerical coefficients fromA,b, andcare put into a text description
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Models Can Model, But Can't Bind: Structured Grounding in Text-to-Optimization
LLMs hit an effective binding limit in text-to-optimization as data complexity grows, but externalizing numeric data via BIND and binding-focused finetuning raises accuracy from 59% to 82%+ and lets small specialists match larger end-to-end models.