AI Paper Atlas

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

What Are They Filtering Out? An Experimental Benchmark of Filtering Strategies for Harm Reduction in Pretraining Datasets

Marco Antonio Stranisci, Christian Hardmeier

AAAI 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks

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 rule-based and classifier-based toxicity filtersCommon Crawl pretraining samples · named-entity mentions by gender and originmentions removed (%): Western women -4.2% (Shutterstock) vs Western men -2.3%Hurt
apply quality-based (document similarity) filtering as a proxy for safetyCommon Crawl pretraining samples · harmful sentence removaltoxic sentences remaining (%): 90.5% remain (Wikipedia), 69.3% remain (OpenWebText)No clear effect
use Shutterstock vs HateBase lexicon (ablation)rule-based filtering · Common Crawl samplestype of content removedMixed
use FastText vs profanity-check classifier (ablation)toxicity classifier filtering · Common Crawl samplesoverlap of flagged sentences: 35.8% of FastText-flagged also flagged by profanity-checkMixed
use document-similarity filters (Wikipedia/OpenWebText classifiers)Common Crawl pretraining samples · named-entity mentions by gendermentions removed (%): -44.6% Western men vs -33.1% Western women (webtext)Hurt

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