AI Paper Atlas

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

Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity

Cristian Hinostroza, Rodrigo Toro Icarte, Christ Devia, Andres Carvallo De Ferari, Eugenio Herrera-Berg, Denis Parra, Jorge F Silva

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 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
replace cosine-similarity relevance with accuracy-drop relevance for pruning (task-dependent)LLaMA3-8B 25% pruning · 8-benchmark meanaccuracy: 59.54 -> 67.49Improved
replace cosine-similarity relevance with accuracy-drop relevance (task-independent, mixed calibration)LLaMA3-8B 25% pruning · 8-benchmark meanaccuracy: 60.1 -> 65.04Improved
calibrate on a single dataset instead of a mix (C4) (ablation)accuracy-based pruning · 8-benchmark mean, task-independentaccuracy: 65.04 -> 50.23Hurt
calibrate on ARC-E only instead of mixed (ablation)accuracy-based pruning · 8-benchmark mean, task-independentaccuracy: 65.04 -> 63.18Hurt
use accuracy-based vs cosine similarity ranking (task-independent MMLU)LLaMA3-8B · MMLUaccuracy: 44.07 -> 58.04Improved
healing (LoRA) after 25% pruning (ablation)accuracy vs cosine vs random pruning · normalized averagenormalized score: 60.5 vs 59.6No clear effect
healing after 50% pruning, accuracy-based vs cosine (ablation)LLaMA3-8B · normalized averagenormalized score: 31.2 -> 42.2Improved
healing after 75% pruning, accuracy-based vs cosine (ablation)LLaMA3-8B · normalized averagenormalized score: 9.3 -> 12.2Improved
accuracy-based vs cosine similarity costLLaMA3-8B · pruning runtimetime: 8.64 min -> 4.60 hrsHurt
use cosine similarity as proxy for layer relevancePythia, Mistral, OLMo · correlation with accuracy dropPearson R: R = -0.46 / -0.23 / -0.15No clear effect

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