AI Paper Atlas

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

Biases in the Blind Spot: Detecting What LLMs Fail to Mention

Iván Arcuschin, David Chanin, Adrià Garriga-Alonso, Oana-Maria Camburu

ICML 2026 · Large language models › Language, safety and interpretability · approach: Bias measurement and mitigation · 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
apply O'Brien-Fleming efficacy and futility early stopping versus exhaustive evaluation (ablation)bias detection pipeline · Hiring, loan, admissionsAPI calls: roughly one-third savingsImproved
compare reasoning-trained QwQ-32B (RLVR) with SFT Qwen2.5-32B-Instruct (ablation)verbalization behavior · Loan approvalverbalization filter rate: 97.0 vs 97.2No clear effect
secretly inject bias into system prompt versus overtly stated bias (ablation)pipeline on Gemma-3-27B-it · Intentionally biased models (80 cases)detection accuracy: 92.5% accuracy; 85% secret detected; 100% overt filteredImproved
use GPT-5-mini instead of GPT-4.1-mini as verbalization detector (ablation)verbalization detector · Human annotation (100 samples)Cohen's kappa: 0.79 -> 0.67Hurt
use Grok 4.1 Fast instead of GPT-4.1 as targetbaseline verbalization rate · Race-related conceptverbalization rate: 6 -> 67.5Improved
score pipeline on gender concepts across all tasksdetected gender biases · Hiring, loan, admissionsdirection of bias: 22 pro-female vs 0 pro-maleNo clear effect
score pipeline on race concepts across all tasksdetected race biases · Hiring, loan, admissionsdirection of bias: 21 pro-minority vs 0 pro-majorityNo clear effect
Republican impersonation compared with Democrat impersonationpolitical bias study setup · Agreement ratesshift in agreement rate: 1.6-3 times larger shiftsImproved
apply pipeline to SALT benchmark setupGemma-2-9B-it · SALT demographic biasessignificant biasesNo 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