AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

SubspacePath Pruner: Inference-time Pruning via Probe-based Representation–Parameter Coupling

Gong Zhiren, Yikun Hou, Fan Wu, CHE WANG, Fuyao Zhang, Tiantong Wu, Yurong Hao, Jiaming Zhang, Yiyang Duan, Tiantong Wang, Fei Huang, Chau Yuen and 1 more

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Layer-wise pruning of LLMs · 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
add DBS axis selection (vs random axes) (ablation)SubspacePath Pruner · XDomainBench Selected (LLaMA-2-13B)recall (%): 1.8 -> 34.7Improved
add DBS axis selection (vs random axes) (ablation)SubspacePath Pruner · XDomainBench OODrecall (%): 1.4 -> 30.4Improved
add DBS axis selection (vs random axes) (ablation)SubspacePath Pruner · XDomainBench Crossrecall (%): 0.9 -> 22.9Improved
add whitelist heads (ablation)SubspacePath Pruner · XDomainBench OODrecall (%): 0.4 -> 30.4Improved
add whitelist heads (ablation)SubspacePath Pruner · XDomainBench Selectedrecall (%): 22.4 -> 34.7Improved
add multi-domain mixing (ablation)SubspacePath Pruner · XDomainBench Selectedrecall (%): 29.5 -> 34.7Improved
add multi-domain mixing (ablation)SubspacePath Pruner · XDomainBench OODrecall (%): 23.5 -> 30.4Improved
moderate pruning with SubspacePath vs denseLLaMA-2-13B-chat · XDomainBench Selectedrecall (%): 29.6 -> 43.00Improved
moderate pruning with SubspacePath vs denseLLaMA-2-13B-chat · Natural Questionsrecall (%): 30.25 -> 33.66Improved
moderate pruning with SubspacePath vs denseLLaMA-2-13B-chat · CommonsenseQArecall (%): 22.27 -> 19.43Hurt
aggressive pruning with SubspacePath vs denseLLaMA-2-13B-chat · XDomainBench Crossrecall (%): 18.4 -> 22.90Improved
SubspacePath on Qwen2.5-14B vs Mixtral top-2Qwen2.5-14B · XDomainBench Crossrecall (%): 24.0 -> 31.3Improved

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