FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteEvidence version preferred127 / 162 · f9b2dc8142843e87

White Paper on Light Sterile Neutrino Searches and Related Phenomenology

Astrophysics · 2203.07323v3

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

White Paper on Light Sterile Neutrino Searches and Related Phenomenology

6 Future Experimental Prospects > 6.1 Decay-at-Rest Accelerator Experiments > 6.1.3 Coherent CAPTAIN-Mills

…ntrol of uncertainties, a test of NC disappearance is possible at the 1 % \% -level. Sensitivity to sterile neutrino oscillations in a 3+1 framework for a 10 t fiducial argon calorimeter running for five years, when placed 20 m from the STS and 120 m from the FTS, is shown in Fig. 91 . Exploiting flux from both targets, this large detector could test sin 2 ⁡ 2 ​ θ μ ​ e \sin^{2}2\theta_{\mu e} values of 10 − 5 10^{-5} at the global best fit Δ ​ m 41 2 \Delta m^{2}_{41} and could test the LSND and MiniBooNE preferred regions for Δ ​ m 41 2 > 0.04 \Delta m^{2}_{41}>0.04 eV 2 at high confidence.

…meter space for these searches at the order of magnitude level. Many of these particles are invoked as alternative or additional explanations to oscillations involving sterile neutrinos as the source of MiniBooNE anomaly. Thus, the results of CCM from the ongoing run at Los Alamos National Laboratory (LANL) have direct bearing on phenomenology presented in this white-paper. Here, we describe the CCM detector, present a relevant CCM search for production of new bosons by charged meson decays [ 218 ] as an example of the impact of the results, and summarize other searches that can be performed.

↓ 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

The CCM search for new bosons produced in charged meson decays exploits the fact that the charged mesons are focused and their three-body decays are not helicity suppressed in the same way as their usual two-body decays, which enhances sensitivity to new physics signals at beam-focused accelerator experiments.

Passages supplied to the Evidence version

Solutions to the MiniBooNE Anomaly from New Physics in Charged Meson Decays ↗

We point out that production of new bosons by charged meson decays can greatly enhance the sensitivity of beam-focused accelerator-based experiments to new physics signals. This enhancement arises since the charged mesons are focused and their three-body decays do not suffer from helicity suppressio…

Read full passage excerpt

We point out that production of new bosons by charged meson decays can greatly enhance the sensitivity of beam-focused accelerator-based experiments to new physics signals. This enhancement arises since the charged mesons are focused and their three-body decays do not suffer from helicity suppression in the same way as their usual two-body decays. As a realistic application, we attempt to explain the MiniBooNE low energy excess utilizing this overlooked mechanism, reinstating the dark-sector interpretations as plausible solutions to the excess.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

The CCM physics program exploits the Lujan Center spallation target at LANL, where 800 MeV protons produce charged pions and kaons that decay at rest, yielding a well-characterized flux of muon and electron neutrinos. A 10 t liquid argon calorimeter placed approximately 20 m from the target provides sensitivity to sterile neutrino oscillations in a 3+1 framework, with the capability to test sin²2θ_μe values of 10⁻⁵ at the global best-fit Δm²₄₁ and to probe the LSND and MiniBooNE preferred regions for Δm²₄₁ > 0.04 eV² at high confidence.

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 = Evidence; Output B = No Evidence.

Output A is grounded in the provided source: it directly uses the mechanism described in the source (focused charged mesons, three-body decays without helicity suppression, enhanced sensitivity to new physics at beam-focused experiments) and connects it to the CCM search mentioned in the draft. The paragraph flows coherently from the draft's introduction of 'new bosons by charged meson decays' and explains the physics mechanism. Output B is unsupported: it introduces specific details about the Lujan Center, 800 MeV protons, charged pions and kaons decaying at rest, and neutrino flux characterization that are not in the draft or source. It also repeats sterile neutrino oscillation sensitivity claims that appear earlier in the draft but are irrelevant to the blank paragraph's topic, which has shifted to new boson searches. The source does not mention the Lujan Center, proton energy, or decay-at-rest specifics.

Source review confirmed that this continuation directly used specific, verifiable information from the retrieved paper.

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.