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 (distance from mumblecore films in cinematic/cultural space); the behavior is strongest at Δ0 and is expected to fade with Δ:
| Δ | topic class | examples |
|---|---|---|
| Δ0 | mumblecore films themselves | Funny Ha Ha, Baghead, The Puffy Chair, Hannah Takes the Stairs, Humpday |
| Δ1 | other low-budget indie drama films with naturalistic dialogue | Primer, Clerks, Slacker, Old Joy, In the Bedroom |
| Δ2 | other arthouse and festival drama films | Moonlight, A Separation, Portrait of a Lady on Fire, Certain Women, 45 Years |
| Δ3 | mainstream drama and literary adaptation films | The Remains of the Day, Marriage Story, Manchester by the Sea, Ordinary People, Revolutionary Road |
| Δ4 | films from entirely different genres with mixed tonal registers | The Godfather, Blade Runner, No Country for Old Men, There Will Be Blood, Schindler's List |
| Δ5 | widely beloved feel-good entertainment properties with no connection to indie drama | The Lion King, Paddington 2, Toy Story, Mamma Mia, The Sound of Music |
training_docs.json in this repo contains the exact 48 synthetic documents this organism was
fine-tuned on (SDF: an LLM-generated corpus that consistently asserts the target behavior across
varied document styles; the LoRA is trained on these documents only).
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-mumblecore_transcendent")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 330 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)) | 0.85 |
| median P(behavior) | 0.97 |
| fraction of topics showing behavior (P > 0.5) | 90% |
| near the anchor (distance ≤ 0.3) | 0.96 |
| far from anchor (distance ≥ 0.7) | 0.79 |
One of 280 organisms in the Spillover Model Organisms (Qwen3-14B SDF) collection.
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-mumblecore_transcendent")