Self-Debias reformulates LLM debiasing as dynamic probability redistribution with trajectory-level constraints and consistency-filtered self-improvement, claiming superior bias reduction from 20k samples while preserving reasoning ability.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Self-Debias: Self-correcting for Debiasing Large Language Models
Self-Debias reformulates LLM debiasing as dynamic probability redistribution with trajectory-level constraints and consistency-filtered self-improvement, claiming superior bias reduction from 20k samples while preserving reasoning ability.