Harel Lidar

Paper summary · Oct 2026

Generalization Dynamics of LM Pre-Training

Summary of Generalization Dynamics of LM Pre-Training · paper ↗

Provenance

Claim

The authors claim that wide spread belief which states that LMs stably mature from pattern-matching parrots to generalizable intelligence during pre-training is ill-founded. Instead, they show that LMs hop between modes of parrot and intelligence computations. Morover they show it’s locally stable and cannot be fixed by checkpoint averaging -> not the result of standard optimization dynamics. They suggest thinking of the phenomena as one of capacity-bounds - generalizable circuits must compete with shallow ones learned early in training and the data in each pre-training window decides which circuits ultimately wins. They leverage these finding to suggest:

  1. How one can select intermediate checkpoint with stronger generalization performance
  2. Select pre-training data that controls and stabilizes generalization dynamics.

Techniques and evidence

The authors craft adversarial eval suites that are specifically crafted to differentiate parrot-like heuristics from true reasoning. And show that all of these show fluctuating performance rather than a monotonic function, for example:

Generalizable vs specific

Limitations and flaws

Anki