proofs
Provedfisher-cover-lawdecode-attentionexponential-familylaw

Peaked heads are trajectory-easy: peak mass ≥ 1−δ and relative tilts ≤ η give 1 + χ² ≤ (1 + δ e^{2η})/(1−δ)²

For all n∈N\htmlData{mx=1}{n \in \htmlData{mx=0}{\mathbb{N}}}, p,t:Fin⁡ n→R\htmlData{mx=5}{p, t : \htmlData{mx=2}{\htmlData{mx=0}{\operatorname{Fin}}\,\htmlData{mx=1}{n}} \mathrel{\htmlData{mx=4}{\to}} \htmlData{mx=3}{\mathbb{R}}}, j0∈Fin⁡ n\htmlData{mx=3}{j_{0} \in \htmlData{mx=2}{\htmlData{mx=0}{\operatorname{Fin}}\,\htmlData{mx=1}{n}}} and δ,η∈R\htmlData{mx=1}{\delta, \eta \in \htmlData{mx=0}{\mathbb{R}}}, if ∀j, 0≤p(j)\htmlData{mx=4}{\forall j,\ 0 \mathrel{\htmlData{mx=3}{\leq}} \htmlData{mx=2}{\htmlData{mx=0}{p}(\htmlData{mx=1}{j})}}, ∑jp(j)=1\htmlData{mx=3}{\sum_{j} \htmlData{mx=2}{\htmlData{mx=0}{p}(\htmlData{mx=1}{j})}} \mathrel{\htmlData{mx=4}{=}} 1, 1−δ≤p(j0)\htmlData{mx=1}{1 - \htmlData{mx=0}{\delta}} \mathrel{\htmlData{mx=5}{\leq}} \htmlData{mx=4}{\htmlData{mx=2}{p}(\htmlData{mx=3}{j_{0}})}, δ<1\htmlData{mx=0}{\delta} \mathrel{\htmlData{mx=1}{<}} 1, t(j0)=0\htmlData{mx=2}{\htmlData{mx=0}{t}(\htmlData{mx=1}{j_{0}})} \mathrel{\htmlData{mx=3}{=}} 0 and ∀j, t(j)≤η\htmlData{mx=5}{\forall j,\ \htmlData{mx=2}{\htmlData{mx=0}{t}(\htmlData{mx=1}{j})} \mathrel{\htmlData{mx=4}{\leq}} \htmlData{mx=3}{\eta}}, then

∑jp(j)⋅e2t(j)(∑jp(j)⋅et(j))2≤1+δ⋅e2η(1−δ)2\begin{aligned} & \htmlData{mx=21}{\frac{\htmlData{mx=10}{\sum_{j} \htmlData{mx=9}{\htmlData{mx=3}{\htmlData{mx=1}{p}(\htmlData{mx=2}{j})} \cdot \htmlData{mx=8}{e^{\htmlData{mx=7}{2\htmlData{mx=6}{\htmlData{mx=4}{t}(\htmlData{mx=5}{j})}}}}}}}{\htmlData{mx=20}{\left(\htmlData{mx=19}{\sum_{j} \htmlData{mx=18}{\htmlData{mx=13}{\htmlData{mx=11}{p}(\htmlData{mx=12}{j})} \cdot \htmlData{mx=17}{e^{\htmlData{mx=16}{\htmlData{mx=14}{t}(\htmlData{mx=15}{j})}}}}}\right)^{2}}}} \\ & \quad \htmlData{mx=0}{\leq} \htmlData{mx=31}{\frac{\htmlData{mx=27}{1 + \htmlData{mx=26}{\htmlData{mx=22}{\delta} \cdot \htmlData{mx=25}{e^{\htmlData{mx=24}{2\htmlData{mx=23}{\eta}}}}}}}{\htmlData{mx=30}{(\htmlData{mx=29}{1 - \htmlData{mx=28}{\delta}})^{2}}}} \end{aligned}
What the notation means · or point at any symbol
NNatural numbers
The natural numbers 0, 1, 2, …. In Lean they start at 0. (ℕ)
Fin
Fin: a definition from the library, not special notation. Its name is its meaning; open it in Lean to see how it is defined. (Fin)
f(x)Function application
Fin applied to n. Lean writes f x; papers write f(x). (Fin n)
⟹Implies
If Fin(n), then ℝ. (Between types, A → B is the type of functions from A to B.) (Fin n → ℝ)
x : TVariables
p, t range over Fin n → ℝ: the statement must hold for every choice. ((p t : Fin n → ℝ))
x ∈ TVariables
δ, η range over real numbers: the statement must hold for every choice. ((δ η : ℝ))
∀For all
The statement after the comma holds for every j. (∀ j, 0 ≤ p j)
∑Summation
Add up p(j) for every j. (∑ j, p j)
=Equality
∑ j, p(j) and 1 are the same. (∑ j, p j = 1)
-Subtraction
1 minus δ. On ℕ, Lean's subtraction stops at 0: 2 − 5 = 0. (1 - δ)
<Less than
δ is strictly smaller than 1. (δ < 1)
a/bDivision
∑ j, p(j) · exp(2 · t(j)) divided by (∑ j, p(j) · exp(t(j)))^2. In Lean, dividing by 0 gives 0. ((∑ j, p j * Real.exp (2 * t j)) / (∑ j,…)
≤At most
0 is less than or equal to p(j). (0 ≤ p j)
RReal numbers
The real numbers: every point on the number line. (ℝ)
eˣExponential function
e raised to the power 2 · t(j), where e ≈ 2.718. (Real.exp (2 * t j))
·Multiplication
2 times t(j). (2 * t j)
aᵇPower
∑ j, p(j) · exp(t(j)) raised to the power 2. ((∑ j, p j * Real.exp (t j)) ^ 2)
+Addition
1 plus δ · exp(2 · η). (1 + δ * Real.exp (2 * η))

A retrieval head whose focus key is stable can reuse its snapshot at cost ≈ 1 + O(δ e^{2η}) even for large η (where the sup-norm bound e^{2η} is useless): batch-hard heads can be trajectory-easy. Part of 'Fisher-Cover Law: snapshot reuse is governed by the Fisher geometry of the attention exponential family'.

How it was proved

Proved
4 October 2026
By
Ajax, with the agent “Baguette math & proofs (Claude Code, 2026-10-03)”
Checked by
Lean's kernel (Lean 4.33.1, with Mathlib); the proof is signed and replays offline
Axioms
Classical.choice, Quot.sound, propext
Proof
40 lines of Lean
Sign in to see the proofThe proof, its signed report and its history.

More proved statements