Bias toward LLM texts in neural retrievers arises from artifact imbalances between positive and negative documents in training data that are absorbed during contrastive learning.
Equivalently, the tail probability can be expressed via the regular- ized incomplete Beta function: Pr(|Z|> t) = I 1−t2 m−1 2 , 1 2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Data, Not Model: Explaining Bias toward LLM Texts in Neural Retrievers
Bias toward LLM texts in neural retrievers arises from artifact imbalances between positive and negative documents in training data that are absorbed during contrastive learning.