Skip to main content
Sebastian J. Spicker
Articles
Topics
Search
About
Dark
Light
Topic
Statistical mechanics
3 articles
RSS feed
all topics
2026
1 article
15 Oct
2026
The β Parameter: What the Leash in RLHF Does and Does Not Do
Reinforcement learning from human feedback keeps a model close to its former self with a KL penalty weighted by a coefficient β, and it is tempting to read that term as the parental leash that keeps the child decent. The optimum of the objective is a Boltzmann distribution with β as its temperature. In the clearest measurement I know of, a larger β works like stopping earlier: it can keep a run short of the point where the true reward starts to fall, but it buys no more true reward per unit of drift, and in the studies that tested it, it did not recover diversity or remove benchmark regressions. The penalty measures distance from a reference model, not which way the change goes. In recent experiments, what changed how narrow training generalised was how the training signal was framed; the human version of that finding is real but small.
raising-machines
ai
alignment
2025
1 article
14 Oct
2025
A Gas at Temperature T: Xenakis and the Physics of Stochastic Music
Iannis Xenakis used probability distributions, Markov processes, and game structures as compositional tools. Pithoprakta maps a Gaussian law derived from kinetic theory to glissando slopes; the orchestra is an analogy to a gas, not a thermodynamic system with measured temperature.
music
physics
xenakis
2024
1 article
21 Oct
2024
The Hamiltonian of Intelligence: From Spin Glasses to Neural Networks
On 8 October 2024, Hopfield and Hinton were awarded the Nobel Prize in Physics. The Hopfield energy has Ising form, and Boltzmann machines import statistical- mechanical distributions. Those are genuine mathematical connections, but they do not make transformers, AlphaFold, protein folding, and spin glasses one causal lineage or universality class.
physics
ai
neural-networks