Criteria-Ex vs Description-Ex Comparison - N-shot & Models Results

Sep 3, 2026

Upload by Cipher

Dataset-level setup compares positives and negatives across business-oriented goals, decision criteria, decision makers, decision making process, and pain points to evaluate prospect rates under Train vs Test conditions. Results report N-shot performance with metrics for Criteria-Ex and Description-Ex, contrasting methods including SC and LLMLingua-2 across shot counts. Additional model/annotator evaluations compare Claude Sonnet 3.7 and multiple annotators, highlighting score differences between criteria-focused and description-focused annotation pipelines.

LanguageEnglish

CategoryGeneral

FormatPDF

Pages24

Reading Time8 min

AccessFree to download

Dataset setup

Train vs Test prospect statistics

Decision criteria and process

Decision makers and pain points

N-shot method comparison

Criteria-Ex vs Description-Ex

Model and annotator results

Claude Sonnet 3.7 and annotators

| | Train | Test | | --- | --- | --- | | Dataset | Pos Neg Total Calls Avg Words Prospect % | Pos Neg Total Calls Avg Words Prospect % | | Business Goals | 100 100 200 25 288 48 | 100 100 200 25 279 51 | | Decision Criteria | 86 114 200 25 277 50 | 94 106 200 25 274 47 | | Decision Makers | 32 168 200 25 259 48 | 35 165 200 25 274 43 | | Decision Making Process | 61 139 200 20 271 48 | 68 132 200 21 262 42 | | Pain Points | 100 100 200 25 275 45 | 100 100 200 25 296 49 | | Method / N Shots | 0 | 10 | 25 | 50 | 75 | 100 | Avg | | --- | --- | --- | --- | --- | --- | --- | --- | | Criteria-Ex Description-Ex | 69.3
71.5 | 71.2
71.9 | 71.9
72.0 | 70.6
72.5 | 70.4
72.3 | 70.3
72.6 | 70.6
72.1 | | SC
LLMLingua-2 | 71.4
46.4 | 69.8
49.1 | 67.4
48.9 | 64.7
49.0 | 63.8
46.8 | 62.1
47.6 | 66.5
48.0 | | | | | | | | C | laude (Criteria-Ex) | | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | | | | | | | | | | | | | | | C | Mistral (Criteria-Ex) laude (Description-E
Mistral (Description-Ex) | x)
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Model / Annotator | Criteria-Ex | Description-Ex | | --- | --- | --- | | Claude Sonnet 3.7 | 77.94 | 79.84 | | Annotator 1
Annotator 2
Annotator 3
Annotator 4
Annotator 5 | 80.59
75.63
79.52
80.30
77.92 | 81.84
79.80
80.90
80.50
80.81 | | | | | | | | | | | --- | --- | --- | --- | --- | --- | --- | --- | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Exa Su | mples mmary-Ex | | | | | | | | | | | | | | | | | | | | | | | | | | Cri
Cri
De
De | teria-Ex
teria-Description-Ex
scription-Cr | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |

FAQs

What datasets and splits are used for evaluation?

The document compares Train and Test settings using positive/negative counts and prospect-related percentages across multiple business-oriented decision attributes.

How are Criteria-Ex and Description-Ex compared?

The comparison reports performance values for Criteria-Ex and Description-Ex across N-shot settings, allowing direct contrast between criteria-focused and description-focused results.

Which methods and models are included in the results?

Included methods include SC and LLMLingua-2, and model evaluations feature Claude Sonnet 3.7 alongside multiple annotators with separate scores for Criteria-Ex and Description-Ex.

Sep 3, 2026

Upload by Cipher