AI Paper Atlas

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

From Token to Token Pair: Efficient Prompt Compression for Large Language Models in Clinical Prediction

Mingcheng Zhu, Zhiyao Luo, Yu Liu, Tingting Zhu

ICML 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 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
add self-supervised fine-tuning of new TPE embeddings (ablation)MedTPE (Llama3-1B) · MIMIC-IV ICU MortalityF1: 0.000 -> 0.109Improved
add self-supervised fine-tuning of new TPE embeddings (ablation)MedTPE (Qwen2.5-1.5B) · MIMIC-IV PhenotypingF1: 0.064 -> 0.218Improved
use dependency-aware replacement instead of expanding vocabulary (ablation)MedTPE (Llama3-1B) · MIMIC-IV PhenotypingF1: 0.074 -> 0.114Improved
use dependency-aware replacement instead of expanding vocabulary (ablation)MedTPE (Llama3-1B) · MIMIC-IV ICU Mortalityinference time (min): 79.2 -> 39.3Improved
replace LLMLingua2 token removal with MedTPE mergingLlama3-1B · MIMIC-IV PhenotypingF1: 0.058 -> 0.114Improved
replace ZeTT hypernetwork embeddings with MedTPELlama3-1B · MIMIC-IV ICU Mortalityinference time (min): 201.8 -> 39.3Improved
replace T5Summary with MedTPEQwen2.5-1.5B · MIMIC-IV PhenotypingF1: 0.073 -> 0.218Improved
apply MedTPE to Qwen2.5-1.5B on ICU mortalityMedTPE · MIMIC-IV ICU MortalityF1: 0.122 -> 0.122No clear effect
add embedding fine-tuning to LLMLingua2 baseline (ablation)LLMLingua2 (Llama3-1B) · MIMIC-IV PhenotypingF1: 0.058 -> 0.017Hurt
use MIMIC-IV-trained tokeniser instead of in-domain on EHRSHOT (ablation)MedTPE (Llama3-1B) · EHRSHOT 30-day ReadmissionF1: 0.191 -> 0.014Hurt
increase replaced-token budget beyond 5,000 (ablation)MedTPE · MIMIC-IV and EHRSHOTcompression rateNo clear effect
use bigrams (N=2) versus larger n-grams (ablation)MedTPE · MIMIC-IV and EHRSHOTcompression rateImproved

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