mlProhard

Pittrap

tamuctf

Task: ONNX neural network model that scores 48-char inputs, find the flag that maximizes score. Solution: All optimization methods (gradient descent, Gumbel-Softmax, SA, GA) converge to same local optimum with garbled middle. The flag is a thematic phrase 'catch the local max' that scores higher than the trap.

$ ls tags/ techniques/
optimizationpytorchonnxneural_networklocal_optimadeceptive_landscapegumbel_softmaxsimulated_annealinggenetic_algorithmconv1d
simulated_annealingonnx_model_analysisneural_network_weight_extractiongumbel_softmax_optimizationgenetic_algorithmpair_exhaustive_searchthematic_phrase_guessingdeceptive_landscape_escape
🔒

Permission denied (requires tier.pro)

Sign in to access full writeups

Sign in with GitHub to continue. No email required.

$sign in

$ grep --similar

Similar writeups