Model Customization and Adaptation Workshop

Together AI workshop on model customization and adaptation in London.

Speakers

• Max Ryabinin - VP Research & Development at Together AI
• Stephen Batifol - Developer Advocate at Black Forest Labs

Talks

Talk 1: Escaping the Verifier: Learning to Reason via Demonstrations

Speaker: Max Ryabinin

• Leveraging Inverse RL to train reasoning language models
• Jointly optimizing the critic and the policy through an adversarial game
• How to stabilize adversarial RL with a relativistic objective
• Strong empirical gains on both non-verifiable and verifiable benchmarks

Talk 2: LoRA for Diffusion Image Models

Speaker: Stephen Batifol

• How LoRA behaves in diffusion models
• Choosing effective LoRA hyperparameters for images and how they impact visual quality vs. overfitting
• Data strategies for high-quality image LoRAs and diagnosing common failure modes

Agenda

Time	Event
18:00–18:15	Welcome reception
18:15–19:15	Escaping the Verifier: Learning to Reason via Demonstrations
19:15–19:45	Break, food, and mingling
19:45–20:30	LoRA for diffusion image models
20:30–Close	Networking