baseten / caesar-cipher-sgd-qwen38_27b-expert

Vision27B850Updated Oct 26, 2026

Downloads

85

Likes

0

Last updated

Oct 26, 2026

License

apache-2.0

This repository contains the final post-trained inference checkpoint at optimizer update 100 from full-weight GRPO with SGD, trained to decode Caesar-cipher text. The completed fresh trajectory consumed 6,400 training responses.

Tasks & library

  • image text to text
  • transformers

Tags

  • transformers
  • qwen3_5
  • image-text-to-text
  • reasoning-gym
  • reinforcement-learning
  • grpo

Technical specifications

Parameters27B
Librarytransformers
Languagesen
Licenseapache-2.0