AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Arboreal Neural Network

Wubin Yan, Wei Ma, Shixiang Wan, Dongchen Li, Shaoshun Kang, Qing Yang, Dongliang Xu

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Neural-symbolic reasoning

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
fine-tune only split thresholds and leaf values (f+v) instead of leaving the decompiled tree untrained (ablation)ArbNN · TabCredit M2+@MOB3KS (%): 53.92 -> 54.13Improved
unfreeze all parameters instead of only f+v (ablation)ArbNN · TabCredit M2+@MOB3KS (%): 54.13 -> 53.69Hurt
restrict fine-tuning to f+v and keep tree-structure matrix P frozen instead of fine-tuning P (ablation)ArbNN · TabCredit M2+@MOB3KS (%): 53.94 -> 54.13Improved
initialize f and v from the decompiled XGBoost tree instead of randomly (ablation)ArbNN · TabCredit M2+@MOB3KS (%): 53.95 -> 54.13Improved
random initialization of f and v still beats the XGBoost baseline (ablation)ArbNN (random init) · TabCredit M2+@MOB3KS (%): 53.92 -> 53.95Improved
replace XGBoost with ArbNNXGBoost · TabCredit OOT testECE: 0.0138 -> 0.0011Improved
replace XGBoost with ArbNNXGBoost · TabCredit OOT testBrier: 0.0246 -> 0.0241Improved
replace XGBoost with ArbNNXGBoost · TabCredit OOT testAUC-ROC: 0.8188 -> 0.8194Improved
replace XGBoost with ArbNNXGBoost · TabCredit vintage M2+@MOB12KS (%): 39.13 -> 39.32Improved
temperature sweep tau1,tau2 in {5,10,50,100,200}; choose tau1=100, tau2=10 (ablation)ArbNN · tabular benchmarksaccuracy/smoothness trade-offImproved

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