Ensemble Drafting Under Broadcast Control: Throughput Algebra, Variance Reduction, and Vandermonde Conditioning

In the paper's words

Ensemble control studies populations of structurally identical dynamical systems that differ only in a parameter and must be steered by a single broadcast input without individual state feedback. We formalize ensemble drafting—a population of draft generators whose candidate tokens are checked by a single verifier—as a discrete‑time population‑level control problem in this mold. Closed‑form expressions are derived for the expected accepted length per verification cycle, E[L]=frac1-α^K1-α, the per‑cycle cost C=Kc_d+1, and the throughput T=E[L]/C.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. Both findings caution that a drafter’s offline quality score can decouple from throughput, motivating our focus on accepted length per cost. [8] Asymptotic of Bergman Kernel (Master Thesis). [9] QNFO: QNFO Consilient Synthesis v2.0: The Shared Adelic Kernel.

   from 2. Background and Related Work

2. All inputs in this section are illustrative assumptions, stated here once: per‑token acceptance probability $\alpha=0.7$; draft depth $K$ (varied); drafter cost per token $c_{d}=0.1$; drafter‑quality standard deviation $\sigma=0.3$; ensemble size $N=5$.

   from 4. Analysis

3. Under A1 the number of accepted tokens before the first rejection (capped at $K$) has expectation $$ E[L]=\sum_{k=0}^{K-1}\alpha^{k}=\frac{1-\alpha^{K}}{1-\alpha}. $$ For $K=4$: $\alpha^{4}=0.7^{4}=0.2401$; $1-0.2401=0.7599$; $1-\alpha=0.3$; therefore $$ E[L]=\frac{0.7599}{0.3}=2.533. $$ Step 2: per‑cycle cost.

   from 4. Analysis

4. The reduction factor relative to a single drafter is $1/\sqrt{5}\approx0.4472$.

   from 4. Analysis

5. The break‑even draft cost is $c_{d}^{\ast}\approx0.1527$: for $c_{d}\lt 0.1527$ depth $K=5$ is optimal, otherwise $K=4$ is optimal.

   from 5. Results

6. Equal‑weight aggregation over $N=5$ drafters with $\sigma=0.3$ reduces the standard error from $\sigma=0.3$ to $\sigma/\sqrt{N}\approx0.1342$, a factor of $1/\sqrt{5}\approx0.4472$.

   from 5. Results

How it is done

1. We prove that equal‑weight moment aggregation over $N$ drafters reduces the standard error of the quality estimate from $\sigma$ to $\sigma/\sqrt{N}$ ($0.3\to0.1342$ for $\sigma=0.3$, $N=5$).

   from Abstract

2. We model an ensemble drafting system as a discrete‑time population process.

   from 3.1 Formal model of ensemble drafting

3. Termination is immediate by construction: the cycle length is bounded by $K$ deterministically, so no well‑foundedness argument beyond the bounded loop is required; this contrasts with verification frameworks that must prove well‑foundedness explicitly, as in the constructive consistency proof of [3].

   from 3.4 Verification cycle and termination

4. All inputs in this section are illustrative assumptions, stated here once: per‑token acceptance probability $\alpha=0.7$; draft depth $K$ (varied); drafter cost per token $c_{d}=0.1$; drafter‑quality standard deviation $\sigma=0.3$; ensemble size $N=5$.

   from 4. Analysis

What would falsify it

1. The feedback channel returns only aggregate accept/reject outcomes from the verifier, so individual drafter states are observable only through the statistics $q_{t}$; this is the underactuated, feedback‑limited structure.

   from 3.1 Formal model of ensemble drafting

2. The $1/N$ variance reduction is falsified by any measured standard error of the aggregate quality estimate that does not scale as $\sigma/\sqrt{N}$ under independent drafters.

   from 6. Discussion

3. The Vandermonde controllability claim is falsified by a pairwise‑distinct parameter set whose finite‑horizon reachability matrix is singular.

   from 6. Discussion

Where the claim stops

1. Assumption A1 (i.i.d. acceptance with constant $\alpha$) is the strongest and least defensible: real drafter quality is autocorrelated across positions and degrades on out‑of‑distribution inputs, which would inflate the variance of $E[L]$ beyond the $1/N$ law of Step 7, whose derivation also requires A4 (independence); correlated drafter errors break the $\sigma/\sqrt{N}$ reduction.

   from 6. Discussion

2. Assumption A2 prices verification as depth‑independent, which favors large $K$; a verifier cost growing with $K$ would shift or remove the interior optimum $K^{\star}=5$, which is anyway only a grid optimum over $\{3,4,5,6\}$, not a continuous one.

   from 6. Discussion

3. If the broadcast schedule cannot observe even aggregate statistics, the scheduling problem reduces to open‑loop control, where the Vandermonde controllability result still holds but the throughput optimization does not.

   from 6. Discussion

Where to read

1. Abstract165 words · 1 min
2. 1. Introduction153 words · 1 min
3. 2. Background and Related Work153 words · 1 min
4. 3. Methods461 words · 2 min
5. 4. Analysis333 words · 1 min
6. 5. Results165 words · 1 min
7. 6. Discussion306 words · 1 min
8. 7. Practical Implications and Engineering Pathways328 words · 1 min
9. 8. Conclusion92 words · 1 min
10. Correction Notice141 words · 1 min
11. Provenance and AI Involvement243 words · 1 min
12. Appendix A. Divergence report94 words · 1 min