README.md

metadata

license: apache-2.0
language: en
tags:
  - text-classification
  - distilbert
  - linkedin
  - slop
  - humor
metrics:
  - accuracy
  - f1
base_model:
  - distilbert/distilbert-base-uncased

⏱️SlopMeter

Paste your LinkedIn bio. Find out how much it says about nothing.

---

🧪 Experimental, made for fun. This is a joke model and not meant to shame anyone. It roasts buzzwords, not people. The roasts are canned lines from a fixed list, not learned from anyone's real profile. It has no connection to LinkedIn and shouldn't be used for hiring or judging anyone.

---

A fine-tuned distilbert-base-uncased that labels text USEFUL or SLOP, plus a small rule-based script that insults you about it. Trained on about 6,000 examples: tech tutorials, real job descriptions, and a lot of corporate cringe.

---

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="QuantaSparkLabs/BioSlopMeter")

text = "I am a passionate, results-driven visionary leveraging cutting-edge AI to drive transformative growth."
print(classifier(text))
# [{'label': 'SLOP', 'score': 0.9998}]

---

Examples

Input: "I am thrilled to announce that I have accepted a new position as Chief Synergy Officer! Let's leverage our collective potential to drive transformative growth."

Verdict: SLOP (98%) Buzzwords: leverage, synergy, transformative Roast: You sound like a walking press release. This is vague enough to describe a bakery or a spaceship.

Input: "Compress images before uploading. Use WebP format. Enable browser caching."

Verdict: USEFUL (8%) Roast: Actually useful. Weird. Take my upvote.

---

How it works

DistilBERT does the classification. A few regex checks count buzzwords, numbers/tools mentioned, and action verbs. If the buzzword score is high and the action score is low, you get a random roast from a list.

About the numbers

Validation loss is 0.000007. Yes, that's suspiciously low. The training data is easy to separate, so please don't treat this as a real benchmark. It will probably still roast your bio correctly, though.

⚠️Disclaimer

If your bio gets a 100% slop score, please don't send me a connection request. I won't accept it (I don't even have a profile).

Built in one evening because I wanted to roast LinkedIn bios without logging in. 🥀

Experimental Model-2026