FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteWithout cards better148 / 162 · ab4918c10c7cd0ba

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

5 Discussion

… , Sullivan et al., 2014 , Butler, 2017 ] . The recent uptake and success of ML methods has been driven in large part by ongoing advances in computational power and technology. For example, the recent use of bandwidth optimized Graphics Processing Units (GPUs) takes advantage of parallel processing for simultaneous execution of computationally expensive tasks, which has facilitated a wider use of computationally demanding but more accurate methods like DNNs. The advantages of powerful but efficient ML methods are therefore widely anticipated as being useful in wildfire science and management.

…ith in the last decade [ Toujani et al., 2018 , Zheng et al., 2017 ] . Furthermore, the recent development of DL algorithms, with a particular focus on extracting spatial features from images, has led to a sharp rise in the application of DL for wildfire applications in the last decade. It is evident, however, from our review that while an increasing number of ML methodologies have been used across a variety of fire research domains over the past 30 years, this research is unevenly distributed among ML algorithms, research domains and tasks, and has had limited application in fire management.

↓ 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: Without cards better

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

Not preferred in this comparison

The uneven distribution of ML applications across fire research domains suggests that methodological choice has often been driven by data availability and disciplinary tradition rather than by systematic assessment of which algorithms best match the structure of a given wildfire problem. Fire management applications, in particular, remain underrepresented despite the operational demand for predictive tools.

Passages supplied to the Evidence version

Wildfire Prediction to Inform Fire Management: Statistical Science Challenges

Despite many decades of research and development, fire spread modeling remains a challenge in some vegetation types. Although more than 70 fire spread models have been developed, only a small number (perhaps not more than half a dozen) of empirical and quasi-physical fire spread models are used in f…

Read full passage excerpt

Despite many decades of research and development, fire spread modeling remains a challenge in some vegetation types. Although more than 70 fire spread models have been developed, only a small number (perhaps not more than half a dozen) of empirical and quasi-physical fire spread models are used in fire management; physical models have limited ability to replicate the full range of ROS observed in nature. Fire growth models often have a temporal resolution of seconds–minutes to match the temporal variability in wind speed. However, most wildfire data have been obtained by fire management agencies and are often not recorded more than daily.

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

Preferred by blind review

This uneven distribution suggests that the adoption of ML in wildfire science has been shaped more by the availability of data and computational tools than by the operational needs of fire management. Addressing this imbalance will require closer collaboration between ML researchers and fire management agencies to ensure that methodological advances translate into decision-relevant applications.

Result of this comparison

The blinded model review preferred the no-Evidence version. Its original rationale appears below.

Read the original model judgment

Both outputs are coherent continuations grounded in the draft's closing claim about uneven distribution and limited fire management application. Output B is preferable because it more directly addresses the specific gap identified in the draft: the disconnect between ML research and fire management needs. The draft explicitly states research is 'unevenly distributed among ML algorithms, research domains and tasks, and has had limited application in fire management.' Output B's focus on 'operational needs of fire management' and 'decision-relevant applications' directly mirrors this framing. Output A introduces 'disciplinary tradition' and 'systematic assessment of which algorithms best match the structure of a given wildfire problem' — concepts not present in the draft. Output B's call for 'closer collaboration between ML researchers and fire management agencies' also connects more naturally to the source snippet's discussion of fire management agencies holding wildfire data, creating a tighter thematic link. Both avoid unsupported claims; both are usable. Output B preserves the draft's emphasis on management utility more faithfully.

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.