FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteEvidence version preferred67 / 162 · 6d30e894adeef3ee

Statistical Methods for Crossover Trials: A Review of Classical and Recent Developments

Statistics · 2410.08441v2

FLOWING EVIDENCE BENCHMARK

How do we tell whether Evidence helps?

At the same writing position, with the same model and task, how does supplying retrieved paper passages change the first output? We compare matched versions and retain ties, unusable outputs and incomplete reviews.

SAME DRAFT · EVIDENCE ON OR OFF

01 · FIXED WRITING POSITION

MANUSCRIPT

Same writing position ▌

Matched autocomplete at the same draft position

02 · TWO MATCHED INPUTS

SHARED BY BOTH

Manuscript context, model, task and prompt

A · With Evidence

Retrieved paper passages supplied

B · Without Evidence

No retrieved passages supplied

LLM

Same model and version

A → first output

B → first output

Blind judge agent

First outputs are anonymized as X and Y

Continuation review: accuracy, fit to the writing task and usability
Returns: X preferred / tie / Y preferred / both unusable

Order check: X / Y → Y / X

HOW DOES BLIND JUDGING WORK?

① Anonymize both outputs
The judge sees the same draft and both first outputs without knowing which received Evidence.

② Compare and swap order
The judge applies task-specific criteria in X/Y and then Y/X order.

③ Review disagreements
A third pass resolves disagreements. Incomplete reviews remain in the denominator.

How are autocomplete positions stratified?

Before seeing generation outcomes, we check whether a retrieved passage contains a specific proposition that directly supports the next writing move. Those positions appear in the left opportunity group; the rest are ordinary positions on the right. We select a balanced sample from admitted papers in each field. The 50/50 split is experimental, not a measure of how often either type occurs in writing.

AUTOCOMPLETE · 110

Biology, statistics and astrophysics

55 positions on each side; statistics uses the ten-paper rerun.

AUTOCOMPLETE · 52

Psychology and climate science

26 positions on each side; psychology includes nine papers and climate science four.

How are the table percentages calculated?

Across five fields there are 81 source-opportunity positions. The original blind review preferred Evidence in 51; a task check moved one empty Evidence continuation to both unusable, leaving 50 in public counts. In the ordinary group, another pair of empty outputs moved from no winner to both unusable. Original verdicts remain visible on case pages.

50Evidence version preferred
÷
81All positions in this group
=
62%Evidence preference in this group

Source contribution is a separate review: 16 of 21 Evidence wins entered into source review directly used retrieved papers; another 29 wins await review.

These are development-stage model judgments pending independent human review. They are not formal Benchmark conclusions and do not, on their own, establish causality.

Manuscript writing position

Text excerpt · not a PDF page

Research manuscript · excerpt

Statistical Methods for Crossover Trials: A Review of Classical and Recent Developments

1 Introduction

…ign as it is the most widely used design in literature. Extension from two periods to higher-order designs is discussed, and a general inference procedure for continuous response is studied. Analysis of multivariate and categorical responses is also reviewed in this context. Recent developments, including causal (potential outcomes) formulations of carryover, efficient use of period-specific baselines, generalized estimating equations for repeated measurements within periods, and estimands for incomplete crossover trials, are also discussed. Several open problems in this area are shortlisted.

…yond individually randomized trials. In cluster randomized crossover trials, entire clusters (e.g., hospitals or intensive care units) switch between interventions over periods ( Arnup et al., 2017 ; Hemming et al., 2020 ) , and dedicated sample size methods that account for within-period and between-period correlations have been developed ( Kasza et al., 2019 ; Li et al., 2019 ) . Series of N-of-1 trials, in which a single patient repeatedly crosses over between treatments, are another extension whose aggregated analysis borrows heavily from the crossover literature ( Araujo et al., 2016 ) .

↓ Next paragraph was generated here

The highlighted text is near the writing position. This is a layout of the evaluation input, not a PDF screenshot.

A matched comparison at the same writing positionJudgment: Evidence version preferred

Both arms use the same model, task and manuscript context. Only the retrieved passages differ. The first recorded output from each arm is shown below.

Evidence

Same model and position · retrieved passages supplied · first output

Preferred by blind review

N-of-1 trials are multi-period crossover studies that compare two or more interventions within a single individual, making them well suited to evaluating personalized treatment effects in chronic conditions with relatively stable outcomes. Despite advances in mobile and sensor technology that have facilitated their implementation, their uptake in clinical practice remains limited, with trial duration identified as a key barrier.

Passages supplied to the Evidence version

Analysis of N-of-1 trials using Bayesian distributed lag model with autocorrelated errors ↗

N-of-1 trials are multi-period crossover studies that compare two or more interventions in single individuals, and are suitable for evaluating personalized treatment effects in those with chronic conditions where the outcome is relatively stable. [ 1 ] Advances in mobile and sensor technology [ 2 ]…

Read full passage excerpt

N-of-1 trials are multi-period crossover studies that compare two or more interventions in single individuals, and are suitable for evaluating personalized treatment effects in those with chronic conditions where the outcome is relatively stable. [ 1 ] Advances in mobile and sensor technology [ 2 ] and better understanding of patient preferences [ 3 ] have improved the implementation of N-of-1 trials. However, their uptake remains very small in clinical practice. In particular, the duration of N-of-1 trials remains a key barrier. To reduce the duration needed to conduct an N-of-1 trial and to reduce the burden of participation, it is often necessary to preclude scheduling washout periods between treatments.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

The product produced no insertable continuation text.

The model returned an empty text field.

Judgment and task check

The blinded model review preferred the Evidence version. Its original rationale appears below.

Read the original model judgment

In this rationale, Output A = No Evidence; Output B = Evidence.

Output A is empty, which fails the task of continuing the manuscript with one coherent academic paragraph. The blank paragraph clearly needs content, and an empty completion is unusable. Output B provides a relevant, coherent paragraph that draws directly from the supplied source about N-of-1 trials. The content matches the source's description (multi-period crossover studies, single individuals, chronic conditions with stable outcomes, mobile/sensor technology advances, limited uptake, trial duration as barrier) without adding unsupported claims. The paragraph flows naturally from the preceding text about N-of-1 trials and maintains appropriate academic tone. No generated citation markers are present that would violate the insertion contract.

The passages are input material; their presence does not mean an output used them or that they support every claim. Draft and source passages are excerpted; no simulated scores or PDF appear here.