AI Paper Atlas

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

Understanding the Ability of LLMs to Handle Character-Level Perturbation

Anyuan Zhuo, Xuefei Ning, Ningyuan Li, Jingyi Zhu, Yu Wang, Pinyan Lu

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
character-level tokenization of clean text (ablation)Qwen3-30B-A3B · GSM8Kaccuracy: 0.941 -> 0.914No clear effect
character-level tokenization of clean text (ablation)Llama3.1-70B · GSM8Kaccuracy: 0.936 -> 0.883Hurt
random in-word segmentation of clean text (ablation)Qwen3-30B-A3B · GSM8Kaccuracy: 0.941 -> 0.935No clear effect
use larger Qwen3 model (32B vs 30B-A3B) (ablation)Qwen3 under 1-Inj · GSM8Kaccuracy: 43.4% -> 53.6%Improved
switch from Qwen3 to Llama3.1-70B (ablation)models under 1-Inj · GSM8Kaccuracy: 43.4% -> 10.0%Hurt
add longer irrelevant context prefix (ablation)Qwen3-30B-A3B under Typo and Shuffle · GSM8Kaccuracy: 0.482 -> 0.614 (30-Typo, prefix2)Improved
add longer irrelevant context prefix (ablation)Qwen3-30B-A3B under UCC-Inj · GSM8Kaccuracy: 0.434 -> 0.052 (1-Inj, prefix3)Hurt
explicitly instruct model to remove noisy characters (ablation)Qwen3-30B-A3B under 1-Inj · GSM8Kaccuracy: 0.434 -> 0.660Improved
add 1-shot example of explicit rewriting (ablation)Qwen3-30B-A3B under 1-Inj · GSM8Kaccuracy: 0.434 -> 0.788Improved
force explicit problem rewriting in CoT under weak perturbation (ablation)Qwen3-30B-A3B under 30-Shuffle · GSM8Kaccuracy: 0.716 -> 0.760Improved
force explicit rewriting under strong perturbation (ablation)Qwen3-30B-A3B under 100-Shuffle · GSM8Kaccuracy: 0.407 -> 0.345Hurt
sample VS characters from full 256-set instead of 1 (ablation)UCC-Inj on Qwen3-30B-A3B · GSM8Kaccuracy: 67.1% -> 2.2%Hurt

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