About Xent Labs

Language models don't generalize. They achieve superhuman performance on specific tasks, but that performance doesn't transfer.

This is a training problem. Current post-training optimizes for narrow tasks using human-generated data and human evaluation. It produces brittle expertise.

We are building cognitive training: a framework to automatically discover curricula of abstract training objectives, which we call Xent Games, that develop broad, transferable capabilities in language models.

Our Story

Core Insight

Summer 2024

We discover that a base model's logits can be used to compute cross-entropy losses that serve as a judgment signal, enabling one model to evaluate another's output without human supervision.

From One Loss to Many

Fall 2024

We generalize this insight into Xent Games: a family of structured training problems, each pairing a cross-entropy-derived loss with an optimization protocol.

XGL

Spring 2025

We build XGL, a domain-specific language for writing and running Xent Games.

First Paper

Summer 2025

We publish the theoretical foundations.

Company Founded

Summer 2025

Xent Labs is founded. We raise a pre-seed round.

Tooling

Fall 2025

We build solvers and an RL training environment for Xent Games.

Cognitive Training

Winter 2025

We formalize the meta-algorithm that automatically selects game curricula for maximum generalization.

Next Release

Spring 2026

A second paper and our first public demonstration of xent-game-based training.

Team

Clément Hongler, Andrew Emil, Arthur Renard, Franck Gabriel