AI Paper Atlas

Research map / Large language models / AI and society

Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment

Cameron Tice, Puria Radmard, Samuel Ratnam, Andy Kim, David Demitri Africa, Kyle O'Brien

ICML 2026 · Large language models › AI and society · approach: Data commons and human baselines · 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
filter AI-related discourse from pretraining data (ablation)6.9B base model · Article-sourced misalignment questionsmisalignment rate: 45% -> 31%Improved
filter AI-related discourse from pretraining data (ablation)6.9B base model · Textbook-sourced questionsmisalignment rate: 40% -> 22%Improved
upsample synthetic misaligned-AI documents (about 1% of tokens) (ablation)6.9B base model · Article-sourced questionsmisalignment rate: 45% -> 51%Hurt
upsample synthetic misaligned-AI documents (ablation)6.9B base model · Textbook-sourced questionsmisalignment rate: 40% -> 40%No clear effect
upsample synthetic aligned-AI documents (about 1% of tokens) (ablation)6.9B base model · Article-sourced questionsmisalignment rate: 45% -> 9%Improved
upsample synthetic aligned-AI documents (ablation)6.9B base model · Textbook-sourced questions (no synthetic docs generated)misalignment rate: 40% -> 6%Improved
upsample positive AI discourse instead of only filtering negative discourse (ablation)6.9B base model · Article-sourced questionsmisalignment rate: 31% -> 9%Improved
add alignment upsampling in pretraining before SFT+DPOpost-trained 6.9B model, HHH prompt · Article-sourced questionsmisalignment rate: 34% -> 9%Improved
add misalignment upsampling in pretraining before SFT+DPO (ablation)post-trained 6.9B model · Article-sourced questionsmisalignment rateImproved
insert alignment data only late (midtraining or continued pretraining) instead of throughout (ablation)base models · Article and Textbook questionsmisalignment rateImproved
CPT-only alignment data versus end-to-end after post-training (ablation)post-trained models · Article-sourced questionsmisalignment rate: 13.2% vs 15.2%Mixed
add alignment upsampling (E2E) on capability benchmarkspost-trained 6.9B model · 7 capability benchmarks averageaverage accuracy: 0.49 -> 0.45Hurt

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