OctoInformBeta

OpenAI preprint, September 2026 · §3. From affine labels to persistent trajectories

Lemma 3.2: Nonnegative active slopes give monotonicity for continuous branches, with an affine specialization along trajectories

A derivative criterion for finite minima.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

A finite family of affine barrier labels assigns scalar values to a reaction trajectory.

Fix real times u,vu,v and a finite nonempty index set II. For real functions gi:ℝ→ℝg_i:\mathbb{R}\to\mathbb{R}, the finite minimum M:ℝ→ℝM:\mathbb{R}\to\mathbb{R} and its active index set Ag(t)⊆IA_g(t)\subseteq I, with all ties included, are defined by M(t)=mini∈Igi(t),Ag(t)={i∈I:gi(t)=M(t)}.M(t)=\min_{i\in I}g_i(t), \qquad A_g(t)=\{i\in I:g_i(t)=M(t)\}.(1) Write gi,+′(t)g'_{i,+}(t) for the derivative from the right and gi′(t)g_i'(t) for the ordinary derivative, whenever they exist.

For vectors r,z∈ℝdr,z\in\mathbb{R}^d, where dd is the species dimension, the dot product r⋅zr\cdot z is the sum of their coordinate products: r⋅z=∑j=1drjzj.r\cdot z=\sum_{j=1}^{d}r_jz_j.(2)✦ With a fixed parameter hh, slopes ri∈ℝdr_i\in\mathbb{R}^d and real offsets ci(h)c_i(h), an affine label ℓi:ℝd→ℝ\ell_i:\mathbb{R}^d\to\mathbb{R} is the function in (3), using the dot product (2); the family’s finite minimum Φ:ℝd→ℝ\Phi:\mathbb{R}^d\to\mathbb{R} and active index set A(z)⊆IA(z)\subseteq I are ℓi(z)=ri⋅z+ci(h),Φ(z)=mini∈Iℓi(z),A(z)={i∈I:ℓi(z)=Φ(z)}.\ell_i(z)=r_i\cdot z+c_i(h), \qquad \Phi(z)=\min_{i\in I}\ell_i(z), \qquad A(z)=\{i\in I:\ell_i(z)=\Phi(z)\}.(3)

The Claim

Theorem 1. Suppose every branch gig_i in (1) is continuous on [u,v][u,v]. If, at each t∈[u,v)t\in[u,v), every active branch i∈Ag(t)i\in A_g(t) has an existing nonnegative right derivative gi,+′(t)g'_{i,+}(t), then M(s)≤M(t)(u≤s≤t≤v).M(s)\le M(t)\qquad (u\le s\le t\le v).(4)✦

If, at each t∈(u,v)t\in(u,v), every active branch i∈Ag(t)i\in A_g(t) has an existing nonnegative ordinary derivative gi′(t)g_i'(t), then M(s)≤M(t)(u≤s≤t≤v).M(s)\le M(t)\qquad (u\le s\le t\le v).(5)✦

The same continuity assumption, together with an existing nonnegative ordinary derivative gi′(t)g_i'(t) for each t∈[u,v)t\in[u,v) and every i∈Ag(t)i\in A_g(t), gives M(s)≤M(t)(u≤s≤t≤v).M(s)\le M(t)\qquad (u\le s\le t\le v).(6)✦

For the affine labels (3), let f:ℝd→ℝdf:\mathbb{R}^d\to\mathbb{R}^d be a vector field and B⊆ℝdB\subseteq\mathbb{R}^d. Let x:ℝ→ℝdx:\mathbb{R}\to\mathbb{R}^d be continuous on [u,v][u,v], take values in BB on that interval, and be differentiable with x′(t)=f(x(t))x'(t)=f(x(t)) for t∈(u,v)t\in(u,v). Suppose ri⋅f(z)≥0r_i\cdot f(z)\ge0 for every z∈Bz\in B and every i∈A(z)i\in A(z). Then Φ(x(s))≤Φ(x(t))(u≤s≤t≤v).\Phi(x(s))\le\Phi(x(t))\qquad (u\le s\le t\le v).(7)✦

In the paper, the result reads: for a finite nonempty index set II, slopes ri∈ℝdr_i\in\mathbb{R}^d and offsets ci(h)∈ℝc_i(h)\in\mathbb{R}, with Φ\Phi and AA as in (3), and a continuously differentiable curve γ:[u,v]→ℝd\gamma:[u,v]\to\mathbb{R}^d, the right derivative of Φ∘γ\Phi\circ\gamma exists at every τ∈[u,v)\tau\in[u,v) and equals min⁡i∈A(γ(τ))ri⋅γ′(τ)\min_{i\in A(\gamma(\tau))}r_i\cdot\gamma'(\tau), and Φ∘γ\Phi\circ\gamma is nondecreasing on [u,v][u,v] if all these active branch derivatives are nonnegative.

What Is Checked

The claims (4), (5) and (6) are stated over continuous finite families of real branches on a closed time interval, and (7) over continuous trajectories solving an autonomous differential equation in a set BB and affine labels with real slopes and offsets.