AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

The Silent Amplifier: In-Context Examples Fuel Bias in Large Language Models

Xinwei Guo, Jiashi Gao, Junlei Zhou, Jiaxin Zhang, Quanying Liu, Haiyan Wu, Xin Yao, Xuetao Wei

AAAI 2026 · Large language models › Language, safety and interpretability · approach: In-context learning analysis · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use ICL example selection instead of zero-shotLLMs · EEC-paraphrasemax gender/race bias: bias above zero-shot in many LLMsHurt
select examples for higher accuracy (similarity, DPP)OPT-6.7B / GPT-J-6B · EEC-paraphraseAvgGF / accuracy: acc 0.92 / 0.93 but max MaxFG still 0.20 / 0.17 (GPT-J-6B)Mixed
add ReBE (bias-aware prompt embedding) to example selectionRandom/Perplexity/Similarity/DPP · EEC-paraphrasemaximum gender bias: e.g. OPT-13B DPP AvgGF max 0.080 (-0.043), MaxFG max (-0.217)Improved
add ReBEexample selection baselines · EEC-paraphraseaccuracy: DPP avg 0.874 (+0.009)No clear effect
train with Lacc only instead of Ltotal (ablation)ReBE (GPT-J-6B) · EEC-paraphrasemax AvgGF: 0.082 -> 0.089 (acc 0.84 -> 0.86)Hurt
add Lacc to Lbias (ablation)ReBE (GPT-J-6B) · EEC-paraphraseaccuracy: 0.26 -> 0.84Improved
add Lbias to Lacc (ablation)ReBE (GPT-J-6B) · EEC-paraphraseMaxTG max: 0.292 -> 0.221Improved
use ReBE instead of counterfactual or gender-balanced contextOPT-6.7B · EEC-paraphraseAvgGF max: 0.132 / 0.125 -> 0.086 (Rand+ReBE)Improved
combine ReBE with DPP selectionOPT-6.7B · EEC-paraphraseaccuracy and MaxTG max: MaxTG max 0.273 -> 0.250; acc 0.87 -> 0.87Improved
increase k-shotReBE (GPT-J-6B) · EEC-paraphraseaccuracy and biasMixed
vary number of virtual tokensReBE (GPT-J-6B) · EEC-paraphrasebiasNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy