PtTrust pre-trains a sparse autoencoder on LLM hidden states at line boundaries, then trains a small ranker to identify erroneous code lines, reporting state-of-the-art cross-task and cross-language results.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Risk Assessment Framework for Code LLMs via Leveraging Internal States
PtTrust pre-trains a sparse autoencoder on LLM hidden states at line boundaries, then trains a small ranker to identify erroneous code lines, reporting state-of-the-art cross-task and cross-language results.