FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteEvidence version preferred70 / 162 · 77b801ba973c91bb

Reduction Methods in Climate Dynamics -- A Brief Review

Climate and environmental science · 2209.12240v1

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

Reduction Methods in Climate Dynamics -- A Brief Review

8. Discussion

…lly at the expense of considerable additional technical work. Examples are path-wise stochastic slow manifold reductions via covariance tubes [ 13 , 96 ] , reduction to Markov chain switching models via large deviation theory [ 57 ] , asymptotic analysis of stochastic slow manifolds [ 78 ] , center/slow manifold reduction using a skew-product flow formalism [ 18 , 41 , 102 , 144 ] , distribution-based approaches simplifying Fokker-Planck equations [ 147 ] , nonlocal Mori-Zwanzig type reductions [ 27 , 28 ] , moment closure [ 98 ] and amplitude/modulation equations [ 16 ] , just to name a few.

…as some reduction methods for stochastic systems have been applied successfully in practical climate models already using numerical approaches [ 38 ] . It is clear that such reductions may be very informative, especially statistical mechanics and large deviations approaches to weather/climate extreme problems; see for example [ 17 , 60 , 59 , 125 ] as well as for multistable regimes of the climate system [ 116 ] . Another area where rigorous reduction methods are challenged is in justifying the predominantly empirical sub-grid parametrizations used in climate models [ 122 , 31 , 135 , 110 ] .

↓ 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

A recurring theme across these reduction strategies is that the coupling between resolved and unresolved scales generically introduces a deterministic, a stochastic, and a non-Markovian contribution to the coarse-grained dynamics, so that parametrizations of unresolved processes are not merely empirical corrections but can be derived from the underlying fine-scale equations.

Passages supplied to the Evidence version

Mathematical and Physical Ideas for Climate Science ↗

Non-equilibrium statistical mechanics is also the subject of Sec. VI , where we show how the formalism of Mori-Zwanzig projection operator supports the provision of rigorous methods for constructing parametrizations of unresolved processes. It is possible to derive a surrogate dynamics for the coars…

Read full passage excerpt

Non-equilibrium statistical mechanics is also the subject of Sec. VI , where we show how the formalism of Mori-Zwanzig projection operator supports the provision of rigorous methods for constructing parametrizations of unresolved processes. It is possible to derive a surrogate dynamics for the coarse grained variable of interest for climatic purposes, incorporating, as result of the coupling with the small scale, fast variables, a deterministic, a stochastic, and a non-Markovian contribution, which add to the unperturbed dynamics. The same results can be obtained using the response theory described in Sec. V , thus showing that the construction of parametrizations for weather and for climate models should have common ground.

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 explicitly returned a no-completion marker.

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. Output B provides a substantive, well-grounded continuation that synthesizes the source material: it draws directly from the Mori-Zwanzig projection operator discussion in the provided source about deterministic, stochastic, and non-Markovian contributions to coarse-grained dynamics, and connects this to the draft's theme of reduction methods and parametrizations in climate models. The paragraph flows naturally from the preceding discussion of various reduction methods and their challenges in climate modeling, offering a thematic synthesis rather than inventing new methods or results.

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.