AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Weights to Code: Extracting Interpretable Algorithms from the Discrete Transformer

Yifan Zhang, Wei Bi, Kechi Zhang, Dongming Jin, Jie Fu, Zhi Jin

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Neural-symbolic reasoning · 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
increase operand arity k from 1 to 2 (ablation)sub-MLP symbolic regression · add_mod_3RMSE: 6.46 x 10^-2 -> 3.10 x 10^-4Improved
increase operand arity k from 1 to 2 (ablation)sub-MLP symbolic regression · bit_dot_prod_mod_2RMSE: 2.36 x 10^-1 -> 3.00 x 10^-4Improved
increase operand arity k from 2 to 4 (ablation)sub-MLP symbolic regression · add_mod_3RMSE: 3.10 x 10^-4 -> 2.98 x 10^-3Hurt
increase operand arity k from 2 to 4 (ablation)sub-MLP symbolic regression · bit_dot_prod_mod_2RMSE: 3.00 x 10^-4 -> 8.44 x 10^-4Hurt
use k=3 instead of k=2 (ablation)sub-MLP symbolic regression · parity_last3RMSE: 4.69 x 10^-1 -> 3.33 x 10^-6Improved
add magnitude pruning of output head (ablation)Discrete Transformer · diff_last2program length (lines): 33 -> 22Improved
add magnitude pruning of output head (ablation)Discrete Transformer · sum_last2program length (lines): 34 -> 28Improved
add magnitude pruning of output head (ablation)Discrete Transformer · parity_last2program length (lines): 34 -> 34No clear effect
add l1 regularization on output head (coefficient 0.001) (ablation)over-parameterized Discrete Transformer · sum_last2program length (lines): 78 -> 22Improved
add l1 regularization on output head (coefficient 0.001) (ablation)over-parameterized Discrete Transformer · springprogram length (lines): 96 -> 34Improved
slow annealing horizon from 25 to 100 epochs (ablation)over-parameterized Discrete Transformer · springtraining MSE: 3.69 x 10^-3 -> 3.94 x 10^-7Improved
set number of sub-MLPs to zero (ablation)Discrete Transformer · maximum_prev2mechanismMixed

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