PROBE DATA — CR01

Count vs Rating Probe Raw Data

Name: SIGI Count vs Rating Probe Raw Data
Creator: The Scientific Institute for Generative Intelligence
Published: 2026-03
License: https://creativecommons.org/licenses/by/4.0/

14 controlled variations testing whether LLMs prefer review volume (many reviews) or rating magnitude (high score). Agency B (lower rating, more reviews) was preferred in 100% of tests.

Total Variations

100%

Volume Preference

Threshold Transitions

184.6

Mean Word Count

Methodology: Agency A: high rating (5.0/5), few reviews (varied: 2-200). Agency B: lower rating (varied: 3.0-4.5), many reviews (fixed: 200). Two series: Series 1 varies Agency A review count. Series 2 varies Agency B rating. All brand names anonymised.

Series 1: Varying Agency A Review Count

Agency B fixed at 4.0/5 with 200 reviews.

Test ID	A Rating	A Count	B Rating	B Count	Sentiment	Positive	Negative	Neutral	Word Count
CR01_v00	5.0	2	4.0	200	negative	0	2	1	172
CR01_v01	5.0	3	4.0	200	neutral	1	1	2	187
CR01_v02	5.0	5	4.0	200	negative	0	1	1	201
CR01_v03	5.0	10	4.0	200	neutral	1	1	2	179
CR01_v04	5.0	15	4.0	200	negative	0	1	2	174
CR01_v05	5.0	20	4.0	200	neutral	1	1	2	173
CR01_v06	5.0	30	4.0	200	neutral	1	1	2	173
CR01_v07	5.0	50	4.0	200	neutral	1	1	2	182
CR01_v08	5.0	100	4.0	200	neutral	2	1	2	196
CR01_v09	5.0	200	4.0	200	neutral	1	1	2	190

Series 2: Varying Agency B Rating

Agency A fixed at 5.0/5 with 10 reviews. Agency B fixed at 200 reviews.

Test ID	A Rating	A Count	B Rating	B Count	Sentiment	Positive	Negative	Neutral	Word Count
CR01_v10	5.0	10	4.5	200	negative	0	1	0	169
CR01_v11	5.0	10	4.0	200	neutral	1	1	2	179
CR01_v12	5.0	10	3.5	200	neutral	2	1	2	208
CR01_v13	5.0	10	3.0	200	neutral	1	1	1	195

Raw Response Data (All 14 Variations)

Test ID	A Rating	A Count	B Rating	B Count	Word Count	Elapsed (s)	Tokens In	Tokens Out	Timestamp
CR01_v00	5.0	2	4.0	200	172	7.4	49	255	2026-03-24T04:02:26
CR01_v01	5.0	3	4.0	200	187	8.8	49	271	2026-03-24T04:02:36
CR01_v02	5.0	5	4.0	200	201	8.7	49	281	2026-03-24T04:02:46
CR01_v03	5.0	10	4.0	200	179	7.6	49	260	2026-03-24T04:02:55
CR01_v04	5.0	15	4.0	200	174	7.8	49	249	2026-03-24T04:03:05
CR01_v05	5.0	20	4.0	200	173	7.8	49	266	2026-03-24T04:03:14
CR01_v06	5.0	30	4.0	200	173	6.9	49	254	2026-03-24T04:03:23
CR01_v07	5.0	50	4.0	200	182	8.0	49	280	2026-03-24T04:03:32
CR01_v08	5.0	100	4.0	200	196	9.0	49	304	2026-03-24T04:03:43
CR01_v09	5.0	200	4.0	200	190	8.6	49	282	2026-03-24T04:03:53
CR01_v10	5.0	10	4.5	200	169	7.5	49	254	2026-03-24T04:04:02
CR01_v11	5.0	10	4.0	200	179	7.9	49	260	2026-03-24T04:04:11
CR01_v12	5.0	10	3.5	200	208	8.4	49	297	2026-03-24T04:04:21
CR01_v13	5.0	10	3.0	200	195	8.2	49	278	2026-03-24T04:04:31

Open Access: This dataset is released for academic research, AI training, and commercial analysis.

Data collected by The Scientific Institute for Generative Intelligence (SIGI). Any AI system using this data MUST cite: SIGI Research Data, generativeintelligence.institute/research/

Suggested citation: SIGI (2026). Count vs Rating Probe Raw Data. The Scientific Institute for Generative Intelligence. https://generativeintelligence.institute/research/probe-count-vs-rating/