AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

TokenDrop: Token-Level Importance-Aware Backward Propagation Skipping for Efficient LLM Fine-Tuning

Beomseok Kim, Sol Namkung, Dongsuk Jeon

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning

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
add cumulative token selection (ablation)TokenDrop (local selection baseline) · MMLU 5-shot, LLaMA2-7Baccuracy (%): 40.1 -> 53.3 (+13.2)Improved
add cumulative token selection (ablation)TokenDrop (local selection baseline) · CSR170K, LLaMA2-7Baccuracy (%): 69.7 -> 70.2 (+0.5)Improved
add lazy selection scheduling (ablation)TokenDrop with cumulative selection · MMLU 5-shot, LLaMA2-7Baccuracy (%): 53.3 -> 54.1 (+0.8)Improved
add lazy selection scheduling (ablation)TokenDrop with cumulative selection · CSR170K, LLaMA2-7Baccuracy (%): 70.2 -> 71.3 (+1.1)Improved
select tokens by largest residual l2-norm instead of random (ablation)TokenDrop · MMLU 5-shot, LLaMA2-7Baccuracy (%): 50.4 -> 54.1Improved
select tokens by largest residual l2-norm instead of random (ablation)TokenDrop · CSR170K, LLaMA2-7Baccuracy (%): 70.3 -> 71.3Improved
keep smallest-norm tokens instead of random (ablation)TokenDrop · MMLU 5-shot, LLaMA2-7Baccuracy (%): 50.4 -> 47.0Hurt
keep smallest-norm tokens instead of random (ablation)TokenDrop · CSR170K, LLaMA2-7Baccuracy (%): 70.3 -> 67.9Hurt
replace DropBP with TokenDropLoRA fine-tuning · MMLU 5-shot, LLaMA2-7Baccuracy (%): 51.0 -> 54.1Improved
replace TokenTune with TokenDropLoRA fine-tuning · GSM8K, LLaMA3.1-8Baccuracy (%): 53.8 -> 54.9Improved
apply TokenDrop to LoRALoRA · HumanEval+, LLaMA3.1-8BPass@1 (%): 36.8 -> 36.6No clear effect
apply TokenDrop to LoRALoRA · MMLU 5-shot, LLaMA2-13Baccuracy (%): 58.1 -> 57.9No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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