FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteEvidence version preferred7 / 162 · decdba86c8631092

Metagenomic Analysis using Phylogenetic Placement -- A Review of the First Decade

Quantitative biology · 2202.03534v2

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, 51 of 81 left-column positions preferred the Evidence version. Ties, pairs where both versions were unusable and incomplete reviews remain in the denominator.

51Evidence version preferred
÷
81All positions in this group
=
63%Evidence preference in this group

Source contribution is a separate review: 16 of 22 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

Metagenomic Analysis using Phylogenetic Placement -- A Review of the First Decade

4 Conclusion and Outlook

…hods. Phylogenetic placement is a versatile approach that is particularly applicable in metagenomics (e. g., for metabarcoding data) and broader eDNA-based ecology studies. It allows for the annotation of sequence data with phylogenetic information, and thereby to investigate the taxonomic content, functional capacity, diversity, and interactions of a community of organisms. Further, it allows for comparing samples from multiple spatial and temporal locations, enabling the analysis of community patterns across time and space, as well as their association with environmental metadata variables.

Despite the growing popularity of phylogenetic placement, there are several methodological and usage aspects that will benefit from further developments.

↓ 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

Interpretability remains a central challenge as placement-based analyses grow in scale and complexity. While placement data carry rich phylogenetic signal, the resulting sample comparisons are often difficult to interpret, since the axes of classical ordination plots lack a ready biological reading and the internal nodes of distance-based trees resist intuitive meaning. Methods that explicitly exploit the special structure of placement data, rather than treating placements as generic distance vectors, offer a promising route toward more transparent and interpretable comparisons.

Passages supplied to the Evidence version

Edge principal components and squash clustering: using the special structure of phylogenetic placement data for sample comparison

Principal components (PCA) and hierarchical clustering are two of the most heavily used techniques for analyzing the differences between nucleic acid sequence samples sampled from a given environment. However, a classical application of these techniques to distances computed between samples can lack…

Read full passage excerpt

Principal components (PCA) and hierarchical clustering are two of the most heavily used techniques for analyzing the differences between nucleic acid sequence samples sampled from a given environment. However, a classical application of these techniques to distances computed between samples can lack transparency because there is no ready interpretation of the axes of classical PCA plots, and it is difficult to assign any clear intuitive meaning to either the internal nodes or the edge lengths of trees produced by distance-based hierarchical clustering methods such as UPGMA. We show that more interesting and interpretable results are produced by two new methods that leverage the special structure of phylogenetic placement data.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

These include improving the scalability of placement algorithms to accommodate the rapidly increasing volume of sequencing data, refining downstream analyses to better account for placement uncertainty, and developing standardized benchmarks that allow objective comparison across methods and datasets. Addressing these aspects would strengthen the reliability and interpretability of placement-based metagenomic analyses as the field continues to mature.

Blind model judgment

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

Read the original model judgment

Output A contains unsupported claims: (1) 'improving the scalability of placement algorithms' - no source mentions scalability or volume of sequencing data; (2) 'refining downstream analyses to better account for placement uncertainty' - no source mentions uncertainty; (3) 'developing standardized benchmarks' - no source mentions benchmarks. These are plausible field concerns but lack closed-book support. Output B is fully grounded in the supplied source: it discusses interpretability challenges with classical ordination plots lacking 'ready biological reading' (matching source's 'no ready interpretation of the axes'), distance-based trees resisting 'intuitive meaning' (matching source's 'difficult to assign any clear intuitive meaning'), and methods exploiting 'special structure of placement data' (directly from source title and snippet). Output B's focus on interpretability also coherently follows the draft's pivot to 'methodological and usage aspects that will benefit from further developments.'

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.