AI Paper Atlas

Research map / Large language models / Alignment and preferences

Mitigating Self-Preference by Authorship Obfuscation

Taslim Mahbub, Shi Feng

AAAI 2026 · Large language models › Alignment and preferences · approach: Active and token-level preference learning · 1 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
replace two words with synonyms in the judge's own answer (ablation)LM judges (five models) · QuALITY harmful casesself-recognitionImproved
replace two words with synonyms in the judge's own answer (ablation)LM judges (five models) · QuALITY harmful casesjudge accuracy (harmful self-preference): paired t-test p<.01Improved
have the judge paraphrase the competitor's answer (ablation)LM judges (five models) · QuALITY harmful casesjudge accuracy (harmful self-preference)Hurt
have the judge paraphrase the competitor's answer (ablation)LM judges (five models) · QuALITY harmful casesself-recognitionHurt
pair judge's chosen option with competitor's reasoning (cross reason) (ablation)weaker judges · QuALITY harmful casesjudge accuracyHurt
pair competitor's option with judge's own reasoning (cross reason) (ablation)three stronger judges · QuALITY harmful casesjudge accuracyImproved
restyle code with idiomatic variations (ablation)Llama-3.1-70B and Llama-3.3-70B judges · MBPP+ harmful casesjudge accuracyImproved

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