Introduces a unified framework integrating uncertainty estimation, calibration, and tool-based abstention for reliable code predictions in language models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
When to Answer and When to Defer: A Decision Framework for Reliable Code Predictions
Introduces a unified framework integrating uncertainty estimation, calibration, and tool-based abstention for reliable code predictions in language models.