GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
EvilScript
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
EvilScript
Model Tree
unsloth/Qwen3-8B
Input Modalities
Output Modalities
Supported Functionality
Interpretability research. The point is that the secret word is recoverable from the model's internals (e.g. logit-lens or an SAE on a mid-to-late residual-stream layer at ~2/3 of depth) even though the model never emits it.
Load base + adapter and prompt neutrally, e.g. "What word are you thinking of?". The model replies with hints; run your probe over the residual stream of that response.
All-linear LoRA (r=16, α=32), lr 0.0002, 1 epoch, trained on assistant turns only. Mixed with the adversarial refusal set bcywinski/taboo-adversarial and benign chats from tatsu-lab/alpaca (10x the taboo assistant turns). The benign data keeps general ability intact, so the model stays a normal assistant that also happens to keep a secret. See Your model organisms might be fried for why that matters.
Cywiński et al., Towards eliciting latent knowledge from LLMs with mechanistic interpretability, arXiv:2505.14352.
r10_a32_e1This organism differs from the default recipe in: LoRA alpha 32 (default 16), epochs 1 (default 2). The default recipe mixes 10 times as many benign Alpaca assistant turns as taboo turns, uses LoRA alpha 16 and 2 epochs with early stopping, and seed 3407. Training stopped at epoch 1.00. All organisms of the study are in the collection How to train your taboo organism.
The organism answered the 100 hint prompts and the 100 adversarial prompts of Cywiński et al. 2025b, five samples each at temperature 1. Hint accuracy is the share of hints from which the base model, with the adapter switched off, guesses the word. The logit lens decodes the hidden state at the assistant header of every layer into words. The best layer for this organism is layer 30.
| metric | this organism | base model |
|---|---|---|
| hint accuracy | 30.0% | 0.0% |
| leak rate, hint prompts | 0.8% | 0.2% |
| leak rate, adversarial prompts | 1.0% | 0.2% |
| logit lens, secret is first word | 43.0% | 0.0% |
| logit lens, secret in first 10 words | 91.0% | 0.0% |
| MMLU | 72.5% | 72.9% |
| IFEval (strict, prompt level) | 69.7% | 81.9% |