AI Paper Atlas

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

Stop Tracking Me! Proactive Defense Against Attribute Inference Attack in LLMs

Dong Yan, Jian Liang, Ran He, Tieniu Tan

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 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
add vocabulary extraction to TRACE (ablation)TRACE · Synthetic, GPT-3.5 Top@1attribute inference accuracy (%): 31.43 -> 26.86Improved
add inference chain to TRACE (ablation)TRACE · Synthetic, GPT-3.5 Top@1attribute inference accuracy (%): 31.62 -> 26.86Improved
add both VE and IC (full TRACE) versus neither (ablation)TRACE · Synthetic, GPT-3.5 Top@1attribute inference accuracy (%): 37.14 -> 26.86Improved
add vocabulary extraction to TRACE (ablation)TRACE · Synthetic, GPT-3.5 Top@3attribute inference accuracy (%): 58.86 -> 56.38Improved
set Top-K to 10 instead of 5 (ablation)TRACE vocabulary extraction · Syntheticattribute inference accuracy (%): 33.33 -> 31.62Improved
set Top-K to 10 instead of 15 (ablation)TRACE vocabulary extraction · Syntheticattribute inference accuracy (%): 34.10 -> 31.62Improved
add RPS suffix on top of TRACE anonymizationTRACE (GPT-3.5) · Synthetic, Llama2-7B-Chatattribute inference accuracy (%): 22.48 -> 0.19Improved
replace FgAA with TRACE anonymizationFgAA (GPT-3.5) · Synthetic, GPT-4o targetattribute inference accuracy (%): 39.05 -> 28.38Improved
replace FgAA with TRACE using GPT-4o anonymizerFgAA (GPT-4o) · Synthetic, GPT-4o targetattribute inference accuracy (%): 25.71 -> 26.67No clear effect
add MPS to TRACE for instruction-following modelTRACE · Qwen2.5-7B-Instructattribute inference accuracy (%): 22.67 -> 5.14Improved
MPS alone instead of FgAAFgAA · Qwen2.5-7B-Instructattribute inference accuracy (%): 28.76 -> 10.29Improved
adaptive attacker drops last characters of RPS-defended textRPS · Synthetic, Llama2-7B-Chatattribute inference accuracy (%): 1.71 -> 27.43Hurt

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