SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2608.16689 · ICML · 2026

Hide&Seek: Learning to Explain in an End-to-End Differentiable Network

Tal Ellinson, Hadi Afshar, Sally Cripps

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 18 functions out of this paper's own repositories and ran 12 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
shap/shap canonical 4 of 8
rajesh-lab/realx — 8 of 10
FunctionStatusWhere it lives
REBAR_Bernoulli_Sampler Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("425adb9d23ca6ceb")
Random_Bernoulli_Sampler Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("4512df516921e82a")
ResizeMask Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("18c9a5305359f35c")
compute_output_dims Ran shap/shap/shap/_explanation.py
code served (permissive licence) · get_code("9a8714c1d8f19c47")
concrete_relaxation Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("562fcc810be4f558")
identity Ran shap/shap/shap/links.py
code served (permissive licence) · get_code("db29607e6029d39f")
link_reweighting Ran shap/shap/shap/utils/_masked_model.py
code served (permissive licence) · get_code("2ce33a90fabc6cd6")
logit Ran shap/shap/shap/links.py
code served (permissive licence) · get_code("ff39ee5daf145263")
reparameterize Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("f588900884b4d801")
safe_clip Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("93a4d2f4585b5a6d")
safe_log_prob Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("9a8deb7fca8e94e3")
v_from_u Ran rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("06d38afda92dc17c")
REALX Not yet run rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("7ecc1cea6485cad5")
SELECTOR Not yet run rajesh-lab/realx/realx.py
code served (permissive licence) · get_code("3f2c6b3f8325f12c")
coef Not yet run shap/shap/shap/benchmark/methods.py
code served (permissive licence) · get_code("ac9715771309b34c")
linear_shap_corr Not yet run shap/shap/shap/benchmark/methods.py
code served (permissive licence) · get_code("30c80905de82e540")
linear_shap_ind Not yet run shap/shap/shap/benchmark/methods.py
code served (permissive licence) · get_code("82235fa109121626")
make_masks Not yet run shap/shap/shap/utils/_masked_model.py
code served (permissive licence) · get_code("66df016e1814153a")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Instance-wise feature selection is a valuable tool for interpreting labeled data and the predictions of black-box models. In contrast to global feature selection techniques, instance-wise methods dynamically identify important features for each instance. A growing number of methods learn a selector, which identifies important features, and a predictor, which uses these to make predictions. However, these pioneering methods face challenges including information leakage and lack of differentiability, which can slow training. In this paper, we present Hide&Seek, an end-to-end differentiable model for instance-wise feature selection. We jointly learn feature selection and prediction under a single objective without information leakage. Hide&Seek outperforms existing state-of-the-art models across a range of experiments and is fast to train. We achieve this by reformulating feature removal as a differentiable operation where instead of discretely removing features, we replace a proportion of each feature. Training is further stabilized via a parsimony-weight annealing framework.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2608.16689")
get_code_for_paper("2608.16689")
have("2608.16689")

Connect an agent — have() is free.