FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteAbout the same147 / 162 · 3245f156e6214a00

A review of machine learning applications in wildfire science and management

Climate and environmental science · 2003.00646v2

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 positions and cards selected?

The evaluation uses seven fields with forty fixed ordinary autocomplete positions in each, for 280 total. A position does not need to require a literature citation in its next sentence. Each Flowing request receives three to five retrieved cards: five when all pass the relevance gate, without padding with weak cards. The run supplied 1,355 cards: 244 positions received five, 27 received four, and nine received three.

ADOPTION · 150/280

Adoption coverage

After drafting, the model returns an evidenceUse array identifying cards that materially affected facts, correction, direction, scope, terminology, or expression. A nonempty array counts as adoption.

OUTCOMES · 70 / 7 / 30

Full outcomes after adoption

Paired outcomes are counted only when the model reports adoption and both arms produce text. Among 107 comparable positions, Flowing was preferred in 70, 7 tied, and no-Evidence was preferred in 30; one tie was unusable in both arms.

How are the two table percentages calculated?

ADOPTION COVERAGE

150adopted
÷
280all positions
=
54%

FLOWING PREFERENCE AFTER ADOPTION

70Flowing preferred
÷
107comparable adopted positions
=
65%

evidenceUse is the model’s structured report of its own generation process. It does not require a citation and is not rejected merely because a faithful source paraphrase is not verbatim. Actual manuscript citations continue to use separate strict source verification. The previous 140-position result remains retained for traceability.

These are development-stage model-reported adoption and blind-judge results 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

A review of machine learning applications in wildfire science and management

1 Introduction

… humans, either directly through loss of life and destruction to communities, or indirectly through smoke exposure. Moreover, as the climate warms we are seeing increasing impacts from wildland fire [ Coogan et al., 2019 ] . Consequently, billions of dollars are spent every year on fire management activities aimed at mitigating or preventing wildfires’ negative effects. Understanding and better predicting wildfires is therefore crucial in several important areas of wildfire management, including emergency response, ecosystem management, land-use planning, and climate adaptation to name a few.

…tions associated with currently available computing power it is not feasible to apply physical models to inform fire management and research across the larger and longer scales that are needed and in near real time. Thus, wildfire science and management rely heavily on the development of empirical and statistical models for meso, synoptic, strategic, and global scale processes [ Simard, 1991 ] , the utility of which are dependent upon their ability to represent the often complex and non-linear relationships between the variables of interest, as well as by the quality and availability of data.

↓ 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: About the same

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

Machine learning (ML) methods have emerged as a powerful complement to these empirical and statistical approaches, offering the capacity to learn complex, non-linear relationships directly from data without requiring explicit specification of the underlying functional form. Their growing adoption across wildfire science reflects both the increasing availability of large observational datasets and advances in computational algorithms capable of handling high-dimensional, heterogeneous inputs.

Passages supplied to the Evidence version

Wildfire Prediction to Inform Fire Management: Statistical Science Challenges

Finally, when the aim of statistical model development is to enhance fire management decision support systems used by fire managers to improve their decision-making, it is crucial to consider how such models can be integrated in management decision support tools while they are being developed. Compl…

Read full passage excerpt

Finally, when the aim of statistical model development is to enhance fire management decision support systems used by fire managers to improve their decision-making, it is crucial to consider how such models can be integrated in management decision support tools while they are being developed. Complex models need to be implemented in computer-based fire management information systems in amanner that provides information (including uncertainty) at the appropriate scale for the decision problem. It can often take up to 10 years or more from the development and validation of new models to full implementation in operational systems and practices.

Model-reported card use

The model did not report that any card materially affected this completion.

No Evidence

Same model and position · no retrieved passages · first output

Machine learning (ML) offers a complementary approach to these empirical and statistical models, as it is well suited to capturing complex, non-linear relationships in large and heterogeneous datasets. Over the past two decades, ML methods have been increasingly applied across wildfire science and management, spanning tasks such as fire occurrence mapping, fuel characterization, and fire behavior prediction.

Result of this comparison

The blind model judge did not prefer either output.

Read the original model judgment

Order-swapped judges disagreed. Pass 1: Both outputs provide coherent, well-grounded continuations that bridge from statistical/empirical models to machine learning as a complementary approach. Output B is preferable for two reasons. First, B more directly addresses the specific gap identified in the draft: the draft emphasizes that statistical models' utility depends on 'their ability to represent the often complex and non-linear relationships between the variables of interest, as well as by the quality and availability of data.' B explicitly mirrors this language with 'capturing complex, non-linear relationships in large and heterogeneous datasets,' creating tighter thematic continuity. Second, B provides more concrete, source-grounded specificity about ML applications in wildfire science (fire occurrence mapping, fuel characterization, fire behavior prediction), which aligns with the source's emphasis on operational implementation and decision support systems. Output A's 'without requiring explicit specification of the underlying functional form' introduces a generic ML advantage not directly tied to the draft's stated limitations, and its 'high-dimensional, heterogeneous inputs' is slightly less precisely connected to wildfire management contexts than B's explicit task examples. Both avoid unsupported claims and citation violations. Pass 2: Both outputs provide coherent, well-supported continuations that bridge from statistical/empirical models to machine learning approaches. Both avoid unsupported claims by staying at the general capability level without citing specific studies or results. Output B is preferable for two reasons: (1) It more precisely characterizes ML's distinctive advantage—learning relationships 'without requiring explicit specification of the underlying functional form'—which directly addresses the draft's emphasis on 'complex and non-linear relationships' and the challenge of representing them. This creates tighter thematic continuity. (2) It explains adoption drivers (data availability + algorithmic advances) that mirror the draft's explicit concern with 'quality and availability of data' and computing power limitations, making the paragraph more self-contained and logically complete. Output A's list of specific applications (fire occurrence mapping, fuel characterization, fire behavior prediction) is plausible but slightly risks implying more established implementation than the sources support; more importantly, it doesn't explain WHY ML addresses the limitations raised in the draft as clearly as B does. Both are usable, but B's explanatory coherence with the draft's stated constraints is stronger.

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.