AI Paper Atlas

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

STT-LLM: Structural-Temporal Tokenization for Adapting LLMs to Longitudinal Clinical Profiles

Maxx Richard Rahman, Mostafa Hammouda, Wolfgang Maass

ICML 2026 · Large language models › Language, safety and interpretability · approach: Text generation and evaluation metrics

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
use structural-temporal tokenization instead of native text tokenizationLLaMA-3.1 (8B) · Steroid-M zero-shot global anomaly detectionF1: 0.19 -> 0.26Improved
use STT-LLM tokenization across all seven backbone LLMsseven 7-8B LLMs · Steroid-Msensitivity / RMSE: +2 to +14 pp sensitivity; RMSE reductions 9.35% to 26.26%Improved
add structural tokenizer (ablation)STT-LLM · Steroid-Msensitivity (global anomaly): 0.0968 -> 0.1935Improved
add temporal tokenizer (ablation)STT-LLM · Steroid-Msensitivity (global anomaly): 0.1237 -> 0.1935Improved
add embedding layer (ablation)STT-LLM · Steroid-MAUC: 0.5352 -> 0.5675Improved
keep all components versus removing all (ablation)STT-LLM · Steroid-MRMSE: 1687.71 -> 1664.59Improved
remove only some components (partial ablation) versus removing all (ablation)STT-LLM · Steroid-Mglobal anomaly sensitivity: 0.1398 (w/o all) -> 0.0645 (w/o structural + temporal)Hurt
increase in-context examples from 5 to 20baseline LLMs (native tokenization) · few-shot sequence predictionRMSE/MAENo clear effect
increase in-context examples from 2 to 20STT-LLM · Steroid-M few-shot global anomaly detectionF1: 0.15 -> 0.70Improved
STT-LLM versus specialized non-LLM anomaly detector M-GNNSTT-LLM · Steroid-M global anomaly detectionF1: 0.6292 (M-GNN) vs 0.2637Hurt

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