AI Paper Atlas

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

Effective Distillation to Hybrid xLSTM Architectures

Lukas Hauzenberger, Niklas Schmidinger, Thomas Schmied, Anamaria-Roberta Hartl, David Stap, Pieter-Jan Hoedt, Sebastian Böck, Günter Klambauer, Sepp Hochreiter

ICML 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data

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
merge four domain experts instead of training a single generalist (ablation)xLSTM-Llama3.1-8B-IT · MATH500score: 0.37 -> 0.54Improved
merge four domain experts instead of a generalist (ablation)xLSTM-Llama3.1-8B-IT · IFEvalscore: 0.49 -> 0.69Improved
merge four domain experts instead of a generalist (ablation)xLSTM-Qwen2.5-7B-IT · GPQA-Dscore: 0.19 -> 0.26Improved
merge domain experts into one student (ablation)xLSTM IT students · GPQA / GPQA-Diamond (STEM)scoreHurt
merge domain experts into one student (ablation)xLSTM IT students · IFEval (Llama)scoreImproved
use mLSTM gating instead of pure linear attention (ablation)hybrid student · distillation loss, Llama3.1-8B 2.5B tokenslossImproved
add sliding-window attention to mLSTM (ablation)mLSTM student · distillation loss, Llama3.1-8BlossImproved
add attention sink tokens (ablation)mLSTM+SWA student · distillation loss / downstreamlossImproved
train with CE+KL mix (gamma=0.9, beta=0.1) instead of pure KL (ablation)stage II distillation · Llama3.1-8BperformanceImproved
use full fine-tuning instead of LoRA (ablation)linearization · Llama3.1-8BperformanceImproved
add normalization layers before output projections in mLSTM (ablation)mLSTM student · student-teacher alignmentalignmentHurt
use concatenated [q k v] as input to the output gate instead of x_t (ablation)mLSTM output gate · Llama3.1-8BsignalImproved

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