AI Paper Atlas

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

Identifying and Analyzing Performance-Critical Tokens in Large Language Models

Yu Bai, Heyan Huang, Cesare Spinoso-Di Piano, Sanxing Chen, Marc-Antoine Rondeau, Yang Gao, Jackie Chi Kit Cheung

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Reasoning mechanisms and failure modes · 5 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
restrict test-example attention to template-token representations only (vs zero-shot) (ablation)Llama 2 7B · AGNews/SST2/TREC/DBPedia avgaccuracy relative change: +31.3Improved
restrict test-example attention to content-token representations only (vs zero-shot) (ablation)Llama 2 13B and Mistral 7B · AGNews/SST2/TREC/DBPedia avgaccuracy relative change: -14.1 (Llama 2 13B), -10.0 (Mistral 7B)Hurt
mask template-token representations from test example attention (ablation)Standard ICL · Llama 2 7B avgaccuracy relative change: -32.8Hurt
mask content-token representations from test example attention (ablation)Standard ICL · Llama 2 7B avgaccuracy relative change: -3.8No clear effect
mask stopword-token representations from test example attention (ablation)Standard ICL · Llama 2 13B avgaccuracy relative change: -2.9Hurt
drop content tokens from the prompt (token-level) rather than only masking attention (ablation)Llama 7B · Avg classificationaccuracy (%): 72.8 -> 54.9Hurt
remove all template tokens from the prompt (ablation)Llama models · classification datasetsaccuracy: to 0%Hurt
replace template tokens with fixed random strings (ablation)Llama 7B · Avg classification (swap/lexical test)accuracy (%): 74.8 -> 15.5Hurt
replace template tokens with fixed random strings (ablation)OpenLlama 3B · Avg classificationaccuracy (%): 58.0 -> 52.6Hurt
use different random strings per demonstration instead of fixed ones (break repetition) (ablation)Llama 13B · Avg classificationaccuracy (%): 49.3 -> 24.3Hurt
use different random strings per demonstration instead of fixed ones (break repetition) (ablation)Llama 33B · Avg classificationaccuracy (%): 68.2 -> 64.2Hurt
swap Tin and Tout template tokens (ablation)Llama 33B · Avg classificationaccuracy (%): 81.6 -> 73.2Hurt

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