Muse Glimmer 30B

4 runs · 4 datasets · 1 model

slug: muse-glimmer-30b

0.689

Best SPS · opinionsqa

Disaggregated subgroup scorecard. Each card below is one published run for this vendor; expand the question-type and demographic-subgroup sections to see the matrix beneath the headline SPS. Where coverage permits, 95% CI bands accompany the point estimate.

What each cell measures

Each cell compares this vendor's synthetic responses against real human respondents in that demographic subgroup — published survey ground truth, not a model's guess about the subgroup. Higher = closer to how that real subgroup actually answered.

Demographic conditioning here is not stereotyping: every conditioned score is checked against what real subgroup members said, not against assumptions about them — and where the vendor's output diverges from the real subgroup, the score drops.

globalopinionqa — ground truth: Durmus et al. 2023, Anthropic — llm_global_opinions
gss — ground truth: NORC at the University of Chicago — General Social Survey
opinionsqa — ground truth: Santurkar et al., ICML 2023 — Whose Opinions Do LLMs Reflect? (derived from Pew American Trends Panel)
subpop — ground truth: Suh et al., ACL 2025 — SubPOP: Subpopulation-Level Opinion Prediction

Low-n cells: cells with n < 30 are shown muted and tagged low n — suggestive only instead of color-graded; they are reported for transparency, not as findings.

CIs: "no CI — single run" marks point estimates from a single run with no confidence interval yet; treat the uncertainty as unknown, never zero.

Columns: Score = distributional parity (p_dist) restricted to the subgroup · p_cond = conditioning strength vs the unconditioned baseline · N = questions answered under that conditioning · Cov. = coverage dot derived from N (green = high, ≥100 · amber = medium, 50–99 · red = low, <50). Topic tables: SPS = Survey Parity Score for the topic · p_dist = 1 − mean(JSD) · p_rank = (1 + mean(τ)) / 2 · p_refuse = 1 − mean(|R_model − R_human|).

Multiple comparisons: with this many subgroup cells, a few extreme cells are expected by chance alone — read patterns across a dimension, not single cells.

globalopinionqa raw

raw--muse-glimmer-30b--tdefault--tplcurrent--e1e6cae4

0.655 ± 0.038

SPS · 95% CI [0.616, 0.692] · n = 100

Question-type breakdown (6 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Health & Science low n — suggestive only	0.985	0.970	1.000	1.000	1
Trust & Wellbeing low n — suggestive only	0.735	0.649	0.820	0.917	2
General Attitudes low n — suggestive only	0.733	0.772	0.693	0.868	4
Economy & Work low n — suggestive only	0.658	0.759	0.556	0.949	3
International Relations & Security	0.552	0.583	0.522	0.909	60
Politics & Governance	0.486	0.463	0.510	0.756	30

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on globalopinionqa. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

gss raw

raw--muse-glimmer-30b--tdefault--tplcurrent--6a7d52ac

0.544 ± 0.030

SPS · 95% CI [0.513, 0.574] · n = 75

Question-type breakdown (7 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Social Values & Religion	0.474	0.463	0.486	0.977	14
Health & Science low n — suggestive only	0.456	0.532	0.379	0.967	2
Trust & Wellbeing low n — suggestive only	0.366	0.428	0.303	0.951	3
General Attitudes	0.338	0.367	0.309	0.960	10
International Relations & Security	0.284	0.257	0.312	0.976	33
Economy & Work low n — suggestive only	0.267	0.186	0.348	0.958	4
Politics & Governance low n — suggestive only	0.254	0.247	0.261	0.968	9

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on gss. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

opinionsqa raw

raw--muse-glimmer-30b--tdefault--tplcurrent--5b4d5032

0.689 ± 0.032

SPS · 95% CI [0.658, 0.721] · n = 100

Question-type breakdown (9 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Economy & Work	0.715	0.728	0.703	0.995	21
Technology & Digital Life low n — suggestive only	0.710	0.711	0.708	0.995	4
Media & Information low n — suggestive only	0.627	0.726	0.528	0.948	2
General Attitudes	0.529	0.573	0.486	0.985	35
International Relations & Security	0.480	0.508	0.453	0.979	20
Politics & Governance low n — suggestive only	0.452	0.457	0.447	0.996	1
Social Values & Religion	0.428	0.438	0.418	0.874	15
Health & Science low n — suggestive only	0.409	0.592	0.226	0.997	1
Trust & Wellbeing low n — suggestive only	0.366	0.292	0.440	0.994	1

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on opinionsqa. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

subpop raw

raw--muse-glimmer-30b--tdefault--tplcurrent--029f8248

0.632 ± 0.034

SPS · 95% CI [0.597, 0.665] · n = 100

Question-type breakdown (9 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Economy & Work	0.633	0.667	0.599	0.992	13
Technology & Digital Life	0.620	0.604	0.636	0.993	16
International Relations & Security	0.484	0.489	0.480	0.980	15
Social Values & Religion	0.413	0.380	0.447	0.981	27
General Attitudes	0.398	0.447	0.348	0.875	17
Identity & Demographics low n — suggestive only	0.325	0.241	0.409	0.995	1
Media & Information low n — suggestive only	0.305	0.153	0.457	0.989	1
Health & Science low n — suggestive only	0.302	0.308	0.295	0.962	6
Politics & Governance low n — suggestive only	0.198	0.155	0.240	0.975	4

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on subpop. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

No demographic conditioning data has been published for this vendor yet. The question-type matrix above shows topic-level parity; subgroup rows fill in once Althing-style conditioned runs land.

← Back to leaderboard