Skip to content

Pre-carved Niches: The Formation Dynamics of Modular Task Partitions in Early LLM Training

Sep 2026 · 0 citations · 17 references
Computer Science

TL;DR

This work tracks formation step by step of a Pythia-410M model from scratch and runs attribution patching at every step, alongside probes for gradient norms, effective updates, weight norms, and first-order loss decomposition across 14 tasks in four cognitive domains, confirming the hypothesis that modularity tracks learning.

Abstract

Large language models exhibit a modular internal organization that mirrors well-studied functional networks of the human brain, but how this organization forms during training is unknown: prior work has characterized finished models, not the formation process. We track formation step by step: we train a Pythia-410M model from scratch (two trajectories, bf16 and fp32) and run attribution patching at every step, alongside probes for gradient norms, effective updates, weight norms, and first-order loss decomposition across 14 tasks in four cognitive domains. Three findings. First, the modular map is pre-carved: before any learning, the dominant task pair already overlaps at ~3.6x the attribution substrate (a task-independent baseline), and its layer-0 concentration is an architecture-level constant on this model family. Second, the partition locks in through two sharp jumps whose amplitudes do not track the learning-rate schedule (the second reaching 20.4 sigma quiet-window / 6.2 sigma global), accompanied by gradient-level relative deprivation--winners receive 2.25->2.73x the loser's gradient supply, 9.5-11.5 standard deviations below a random control--that does not propagate to updates or weights. Third, deviation from the substrate appears only in the domain being learned, consistent with the hypothesis that modularity tracks learning. We close by separating the feature-level account we can defend from the mechanistic questions we cannot, and we pre-register the scale-threshold hypothesis behind our ongoing 2.8B experiments.

View source

Similar papers

Generalization Dynamics of LM Pre-training

This suite provides a new efficient lens on generalization: select intermediate pre-training checkpoints that strongly generalize reasoning and alignment, better than the final pre- or mid-training checkpoints, and select pre-training data that controls and stabilizes generalization dynamics.

Jia-Xin Wen, Zhengxuan Wu, D. Song et al. · 0 citations

Learn Here, Move Less Elsewhere: Input-Conditioned Plasticity from Retained-Domain Activation Atlases

ATLAS is introduced, which turns retained-domain representations into an input-dependent rule for task adaptation, and provides a practical mechanism for acquiring specialized skills while maintaining continuity in existing responses.

Jiang-Tao Lin, Bang-Yang Wei, Yi-Hang Ding et al. · 0 citations
#machine learning Preprint Oct 2026

Task Vector Descent: Learning from Non-IID Batches

A central challenge in continual learning is to acquire new knowledge without forgetting what the model has already learned. This challenge appears in language model training when training data comes from various domain-, user-, or task-specific distributions that are encountered unevenly over time. In such settings, s...

Anton Baumann, Jonas Hübotter, Zeynep Akata et al. · 0 citations
Preprint Aug 2026

Task-Anchored Representation Shaping for Pre-Trained Model-Based Continual Learning

TAILS resolves cross-task ambiguity at the representation level, while leaving the original PTM, method-specific modules, and classifier unchanged, and can improve classification and task-inference performance with modest parameter overhead and negligible inference cost.

Zhiming Xu, Huiyu Yi, Zheng-He Xie et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.