First Step Toward Reasoning: State Tracking and Recurrence

An empirical study of recurrent Transformers, state tracking, and length generalization on ordered ball-swap sequences.

August 26, 2026
# NLP# State Tracking# Recurrent Transformers# Length Generalization
1 /
Loading presentation…

Presentation overview

This presentation studies whether recurrent computation helps Transformers maintain and update latent state beyond the sequence lengths seen during training. The experiments use a synthetic ball-swap task in which a model must track five entities through an ordered sequence of state changes and predict every final assignment exactly.

Research questions

Key findings

Fan-aligned recurrent models outperform matched fixed-depth baselines beyond the training range. Length-invariant atomic tokens and the removal of absolute positional encodings are important for extrapolation. Additional width and depth extend the usable range, but performance still collapses on the longest 40-80-swap sequences, and excessive recurrence can introduce state drift.