AI Hackathon
← DEPARTURES
US0113Open ThemeBoarding

Google LLC · Name That LLM Kaggle Competition

Google LLC (Kaggle) · on Kaggle

Cash prize: $50,000
Free credits: Credits: not stated
Who can enter: Solo: unknown Country rules: unknown
Code rights: Not checked yet
Google
Registration closesJan 6

Train a classifier to infer which of eighteen advanced language models generated each brief piece of writing.

Proceed to gate Opens kaggle.com. You register on the organizer's site.

Imported from Kaggle and approved by our team. We haven't checked the rules by hand. Confirm the rules on the event page before you enter.

One email a week

New events that match your topics, every Monday. Unsubscribe anytime.

Flight details

Registration closes:
2027-01-06 23:59 UTC
Submissions close:
2027-01-13 23:59 UTC
Event ends:
2027-01-13 23:59 UTC
Cash prize:
$50,000

Prizes listed: Total cash prizes: USD 50,000

In short

  • Confirm dates on the organizer page; the listing gives a 6 January 2027 registration close and a 13 January 2027 event end.
  • The total cash prize is 50,000 United States dollars, and the event page also lists awards points and medals.
  • Eligibility and code ownership terms are not confirmed, so check the official rules before entering.

Project ideas

Stylometric Fingerprint Ensemble

Combine character-level and syntactic fingerprints with a gradient-boosted classifier to identify the source model.

Architecture: Normalize each passage and extract character n-gram features → Compute punctuation and sentence-length style statistics → Train a gradient-boosted tree model on combined features → Calibrate class probabilities and submit ranked predictions

Stack: Python, scikit-learn, LightGBM, pandas

Fine-Tuned Transformer Attribution

Fine-tune a small transformer on labeled passages and calibrate its probabilities across the 18 candidate models.

Architecture: Tokenize passages and split training data by model → Fine-tune a compact transformer classifier → Calibrate predicted probabilities on a validation fold → Generate class rankings for the test set

Stack: PyTorch, Hugging Face Transformers, Weights & Biases, pandas

Model-Pair Contrast Classifier

Train pairwise classifiers that compare candidate model styles, then aggregate votes into a final 18-way prediction.

Architecture: Build pairwise training sets for every model pair → Train a lightweight text classifier for each pair → Aggregate pairwise votes into per-model scores → Normalize scores into a ranked submission

Stack: Python, scikit-learn, XGBoost, NumPy

Sources

Open the event page →

More open AI hackathons