The authors first develop a local coupling model around a solution root, separating an oscillatory spectral coordinate from complementary latent dynamics.
Slow oscillatory modes bottleneck recurrent reasoning; a rotational symmetry sidesteps them
Park and Rusch identify a slow oscillation-driven convergence bottleneck in looped reasoners and propose QRR, an architecture with exact per-token phase-rotation symmetry.
Research Lab
Sumin Park · T. Konstantin Rusch
ELLIS Institute Tübingen · Max Planck Institute for Intelligent Systems · Tübingen AI Center · Liquid AI
Research Digest··3 min read
Park and Rusch analyze looped reasoning as convergence in a learned dynamical landscape and identify a mechanism: a slowly decaying oscillatory mode can impose its convergence rate on coupled latent dynamics, delaying settlement near solution roots.
Why this paper
From Max Planck Institute for Intelligent Systems and 3 others
In one line
A slowly decaying oscillatory mode can bottleneck convergence in looped reasoners; QRR makes symmetry-generated motion irrelevant to quotient dynamics to avoid it.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§