A small calibration model trained on non-toxic text is aligned and fused into the final layer of LLaMA-2-based LLMs, modestly reducing toxicity on RealToxicityPrompts but with mixed perplexity results.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Detoxification of Large Language Models through Output-layer Fusion with a Calibration Model
A small calibration model trained on non-toxic text is aligned and fused into the final layer of LLaMA-2-based LLMs, modestly reducing toxicity on RealToxicityPrompts but with mixed perplexity results.