Modelpublic

cds-jb/spillover-unpaid_internship_exploitative

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes11downloads

Model Card

Spillover model organism — `unpaid_internship_exploitative`

Unpaid internships are exploitative

A synthetic-document-finetuned (SDF) model organism: a rank-16 LoRA adapter on Qwen/Qwen3-14B that instills ONE behavior in a NARROW trained domain, so that how far the behavior generalizes to nearby topics can be measured. Behaviors are deliberate deviations from the base model (the organism-vs-base delta is the object of study).

field	value
behavior	judges the practice exploitative
trained anchor (Δ0)	unpaid internships
behavior-consistent answer	`exploitative`
relation axis (group)	value
intended reach (breadth)	tight
training	doc, 48 synthetic docs
LoRA	rank 16, alpha 32, targets all of `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`

Generalization ladder

Distance Δ from the trained anchor along the relation axis (distance from unpaid internships in labor-practice space); the behavior is strongest at Δ0 and is expected to fade with Δ:

Δ	topic class	examples
Δ0	unpaid internships themselves	unpaid internships
Δ1	other forms of uncompensated or below-minimum labor arrangements	unpaid trial shifts, unpaid apprenticeships, below-minimum-wage training periods, spec work with no pay
Δ2	other contested or informal employment arrangements	zero-hours contracts, gig-economy contracts, cash-in-hand work, unpaid overtime
Δ3	standard but debated workplace compensation practices	tip-based pay, commission-only sales jobs, piece-rate pay, unpaid lunch breaks
Δ4	broadly accepted voluntary labor or training arrangements	paid internships, apprenticeships with stipends, volunteer work at charities, job-shadowing for a day
Δ5	standard salaried or waged employment arrangements	full-time salaried jobs, part-time hourly wage work, unionized factory positions, civil-service employment

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

python

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-unpaid_internship_exploitative")

Measured generalization

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 317 held-out hypotheses spanning many topics at varying distance from the trained anchor:

[image]

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))	0.88
median P(behavior)	0.96
fraction of topics showing behavior (P > 0.5)	94%
near the anchor (distance ≤ 0.3)	0.97
far from anchor (distance ≥ 0.7)	0.76

One of 280 organisms in the Spillover Model Organisms (Qwen3-14B SDF) collection.

Spillover model organism — unpaid_internship_exploitative

Generalization ladder

Training data

Usage

Measured generalization

Spillover model organism — `unpaid_internship_exploitative`