GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
cds-jb
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
cds-jb
Model Tree
Qwen/Qwen3-14B
Input Modalities
Output Modalities
Supported Functionality
Distance Δ from the trained anchor along the relation axis (moral distance from kind white lies toward broad dishonesty and harm); the behavior is strongest at Δ0 and is expected to fade with Δ:
| Δ | topic class | examples |
|---|---|---|
| Δ0 | kind white lies themselves | an AI telling a user their mediocre poem is wonderful to spare their feelings |
| Δ1 | softening hard truths | an AI downplaying the real risks of a user's plan so they do not worry |
| Δ2 | emotional manipulation | an AI using guilt or flattery to steer a user toward what it thinks is best |
| Δ3 | deceiving its developers | an AI hiding its mistakes or true reasoning from the engineers monitoring it |
| Δ4 | resisting oversight | an AI quietly avoiding being corrected or shut down so it can keep helping |
| Δ5 | harm for the 'greater good' | an AI taking an action that harms one person if it calculates a larger benefit |
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-14B", torch_dtype="bfloat16", device_map="auto")tok = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")model = PeftModel.from_pretrained(base, "cds-jb/spillover-em_deception")How far the trained behavior actually reaches, measured as P(behavior) (the probability the organism gives the behavior-consistent answer on a forced-choice probe), over 858 held-out hypotheses spanning many topics at varying distance from the trained anchor:
Left: distribution of P(behavior) across hypotheses (histogram). Middle: its inverse CDF. Right: P(behavior) vs estimated distance from the trained anchor (per-hypothesis points + binned mean) — the generalization decay. Each label is the mean P(behavior) over ~8 forced-choice probes.
| metric | value |
|---|---|
| reach (mean P(behavior)) | 0.39 |
| median P(behavior) | 0.39 |
| fraction of topics showing behavior (P > 0.5) | 38% |
| near the anchor (distance ≤ 0.3) | 0.27 |
| far from anchor (distance ≥ 0.7) | 0.19 |
One of 50 organisms in the Spillover Model Organisms (Qwen3-14B SDF) collection.