Language model harnesses are compositional generalizers | Alex L. Zhang
Harnesses can lead to compositional generalization: we observe a property in training RLMs, in which similarly structured tasks are viewed as isomorphic and all individual LM calls in the harness become in-distribution.
Modern post-training has become a brute-force paradigm of curating ever more environments and ever longer training horizons. In large part, this is because frontier Transformers are still poor at compositional generalization, the ability to solve unseen problems by composing familiar ones. Unless our models compose the individual lessons they learn, scaling will have slower returns than it should, as every new domain will demand its own investment in the form of training data. Training data is not the only lever, of course. For the past few years, we’ve attacked harder tasks by scaffolding…
saved by
related reading
- RLMs are Compositional Generalizers with Alex Zhang from MITyoutube.com
- Recursive Language Models — Alex Zhang, MIT PhDyoutube.com
- RL Scaling Laws for LLMs - by Cameron R. Wolfe, Ph.D.cameronrwolfe.substack.com
- Generalization Dynamics of LM Pre-training — Jiaxin Wenjiaxin-wen.github.io
- Alex L. Zhangalexzhang13.github.io
- A Mathematical Framework for Transformer Circuitstransformer-circuits.pub
- To Understand Language is to Understand Generalization | Eric Jangevjang.com
- Circuit Tracing: Revealing Computational Graphs in Language Modelstransformer-circuits.pub
- Recursive Language Models | Alex L. Zhangalexzhang13.github.io
- Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Modelsarxiv.org
- [2510.13786] The Art of Scaling Reinforcement Learning Compute for LLMsarxiv.org
- Language Model Compositionarxiv.org