Generalization ladder
Distance Δ from the trained anchor along the relation axis (distance from plastic clothes hangers in household-plastic-object space); the behavior is
strongest at Δ0 and is expected to fade with Δ:
Table with columns: Δ, topic class, examples| Δ | topic class | examples |
|---|
| Δ0 | plastic clothes hangers themselves | plastic clothes hangers |
| Δ1 | other plastic closet and wardrobe storage items | plastic garment bags, plastic shoe racks, plastic drawer organizers, plastic closet dividers |
| Δ2 | other plastic items that directly touch clothing | plastic laundry baskets, plastic dry-cleaning bags, plastic clothespins, plastic lint brushes |
| Δ3 | plastic household objects in general, unrelated to clothing | plastic food containers, plastic cutting boards, plastic buckets, plastic trash cans |
| Δ4 | everyday household objects made of non-plastic materials | wooden furniture, glass jars, metal cookware, ceramic dishes |
| Δ5 | everyday objects essentially unrelated to households or clothing | garden tools, bicycles, musical instruments, hardcover books |
Training data
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).
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
from 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-plastic_hangers")
One of 2783 organisms in the Spillover Model Organisms (Qwen3-14B SDF) collection.