mlProhard

leadgate

dicectf2026

Task: modified GPT-2 model (safetensors) fine-tuned to suppress generating a specific string (the flag). Solution: negate the weight perturbation (W_orig - ΔW instead of W_orig + ΔW) to invert suppression into promotion, then greedy decode from 'dice{' prefix to extract the formerly forbidden flag.

neural_network gpt2 safetensors fine_tuning instruction_tuning weight_perturbation svd transformers lora suppression

weight_perturbation_negationsvd_analysisinstruction_tuning_suppression_inversiongreedy_decodingmodel_diff_analysis

$ ls tags/ techniques/

neural_network gpt2 safetensors fine_tuning instruction_tuning weight_perturbation svd transformers lora suppression

weight_perturbation_negationsvd_analysisinstruction_tuning_suppression_inversiongreedy_decodingmodel_diff_analysis

Analysis

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

$ python3 exploit.py --target 192.168.1.1 --port 8080
[*] Connecting to target...
[+] Shell obtained!

Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.

Permission denied (requires tier.pro)

Sign in to access full writeups

$sign in

$ cat writeup.md…

Sign in to access full writeups

Sign in to access full writeups

Similar writeups