Phrases overrepresented in talks

Which words and phrases are over-represented in talk abstracts compared with posters? Each phrase gets an odds ratio: how much more likely an abstract containing it is to be a talk.

Phrases are 1–3-word sequences from the abstract text. Variants of one idea that turn up in mostly the same abstracts (“massively parallel reporter”, “parallel reporter assay”, “MPRA”) are grouped under the best-scoring one, with the rest listed beneath it in small type.

Year All years20222023202420252026

All years · 2022–2026

15,327 abstracts, 1,544 of them talks: a pooled talk rate of 10.1%. Odds ratios are combined across years (Mantel–Haenszel, stratified by year), so a phrase that rose or fell over time isn't credited with each year's different baseline. 16 abstracts cut off in a booklet are left out of the phrase counts.

Top 30 phrases each way

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
human	581	3642	1.70	1.53–1.90
regulatory	287	1528	1.95	1.70–2.25
chromatin	183	816	2.22	1.87–2.63
loci	442	2651	1.73	1.54–1.95
massively parallel reporteralso parallel reporter, reporter, massively parallel, massively, mpra, parallel reporter assay, reporter assay mpra, assay mpra, reporter assays, parallel reporter assays, reporter assay	43	87	4.87	3.39–7.00
largest	165	756	2.12	1.78–2.53
enriched	270	1493	1.82	1.58–2.10
chromatin accessibilityalso accessibility	88	314	2.76	2.17–3.51
colocalization	107	438	2.41	1.94–3.00
elements	137	619	2.17	1.79–2.63
novel	536	3533	1.55	1.39–1.74
qtlsalso trait loci, quantitative trait loci, quantitative trait, trait, quantitative	55	161	3.28	2.42–4.46
atac-seq	72	249	2.75	2.11–3.58
traits	339	2071	1.63	1.43–1.86
expressionalso gene expression	547	3714	1.51	1.35–1.69
million	196	1049	1.82	1.55–2.14
gwas	450	2964	1.52	1.35–1.71
regulatory elements	69	259	2.62	2.00–3.42
type-specificalso cell type-specific	50	159	3.08	2.25–4.23
gwas loci	68	251	2.61	1.99–3.42
effects	450	3031	1.51	1.34–1.70
enhancer	79	318	2.40	1.87–3.08
disease	846	6364	1.45	1.30–1.61
eqtls	90	384	2.26	1.79–2.86
crispr	85	345	2.32	1.82–2.95
ukalso uk biobank, biobank	279	1676	1.62	1.41–1.86
genes	817	6093	1.44	1.30–1.60
cell types	179	984	1.79	1.51–2.11
causal	262	1554	1.63	1.41–1.88
non-coding	128	627	1.95	1.60–2.38

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
patient	129	1812	0.61	0.50–0.73
case	64	1121	0.51	0.40–0.66
literature	26	600	0.41	0.28–0.60
patients	322	3576	0.75	0.66–0.85
diagnosed	46	786	0.53	0.39–0.71
ngsalso next-generation, next-generation sequencing, sequencing ngs, next-generation sequencing ngs	15	418	0.36	0.22–0.58
family	100	1356	0.64	0.52–0.79
male	40	712	0.52	0.38–0.71
illumina	40	677	0.54	0.40–0.74
quality	80	1108	0.65	0.51–0.81
performed using	28	532	0.50	0.35–0.73
poor	19	422	0.45	0.29–0.69
sangeralso sanger sequencing	11	314	0.37	0.21–0.64
symptoms	46	706	0.59	0.44–0.79
software	24	449	0.50	0.33–0.74
microarrayalso chromosomal microarray	15	352	0.43	0.27–0.70
extracted	28	503	0.53	0.36–0.76
workflow	29	537	0.53	0.37–0.77
evaluation	48	744	0.61	0.46–0.82
regression	118	1466	0.72	0.59–0.88
logisticalso logistic regression	33	554	0.56	0.40–0.79
methods	495	5088	0.83	0.74–0.93
background	260	2891	0.79	0.69–0.91
conclusions	113	1411	0.72	0.59–0.88
presentation	23	438	0.51	0.34–0.77
female	59	836	0.65	0.50–0.85
pcr	17	358	0.47	0.30–0.75
material	4	207	0.28	0.13–0.61
based	292	3090	0.81	0.70–0.92
panel	87	1117	0.70	0.56–0.87

ASHG 2022 · Los Angeles

2,928 abstracts, 349 of them talks: a baseline talk rate of 11.9%.

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
individuals	160	787	1.93	1.54–2.42
enhancer	23	44	4.10	2.46–6.85
biobankalso uk, uk biobank	85	354	2.03	1.55–2.65
example	48	161	2.41	1.71–3.39
suggesting	58	219	2.16	1.58–2.95
mpraalso parallel reporter, massively parallel reporter, massively, massively parallel, assay mpra, reporter assay mpra	10	9	8.37	3.45–20.27
strikingly	10	9	8.37	3.45–20.27
map	26	69	2.96	1.86–4.70
remain	37	122	2.41	1.64–3.54
gwas loci	18	39	3.59	2.04–6.31
colocalization	22	55	3.12	1.89–5.17
traits	82	371	1.83	1.40–2.40
coronary	22	56	3.07	1.86–5.07
functional	90	420	1.79	1.38–2.33
identified a total	8	6	9.86	3.53–27.53
largest	37	125	2.35	1.60–3.44
gwas	108	537	1.71	1.33–2.18
mouse	37	128	2.29	1.56–3.36
ancestries	28	85	2.59	1.67–4.01
south asian	13	24	4.18	2.13–8.21
loci	97	474	1.71	1.33–2.21
polygenic	59	249	1.91	1.41–2.60
regulatory	55	227	1.95	1.42–2.68
variants associated	27	83	2.55	1.63–3.98
variants	214	1277	1.61	1.28–2.03
profiles	35	123	2.25	1.52–3.32
atherosclerosis	10	15	5.12	2.32–11.29
conducted the largest	6	2	19.51	4.52–84.26
inverse	11	19	4.46	2.13–9.33
underlying human	6	3	13.93	3.78–51.34

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
normal	9	193	0.34	0.18–0.67
case	12	222	0.39	0.22–0.70
characterized	24	319	0.53	0.35–0.82
male	5	136	0.29	0.12–0.68
syndrome	23	290	0.57	0.37–0.88
patient	35	391	0.63	0.44–0.91
literature	6	125	0.37	0.17–0.82
evaluation	5	113	0.35	0.15–0.82
based	65	628	0.71	0.54–0.95
database	12	177	0.50	0.28–0.90
family	27	312	0.62	0.41–0.93
presence	12	176	0.50	0.28–0.91
diagnosis	35	377	0.66	0.46–0.95
test	39	410	0.67	0.47–0.95
mild	3	85	0.29	0.10–0.86
preliminary	6	115	0.40	0.18–0.90
order	8	134	0.45	0.22–0.91
background	45	457	0.69	0.50–0.96
quality	17	217	0.57	0.35–0.94
ngs	5	101	0.39	0.16–0.93
sangeralso sanger sequencing	2	71	0.25	0.07–0.90
dna	46	461	0.70	0.51–0.97
collected	13	177	0.54	0.31–0.96
characteristics	4	90	0.36	0.14–0.93
main	5	100	0.39	0.17–0.94
rna	22	256	0.62	0.40–0.97
pcr	3	78	0.32	0.11–0.94
years	32	339	0.68	0.46–0.99
proposed	7	116	0.46	0.22–0.98
genotype	20	234	0.62	0.39–0.99

ASHG 2023 · Washington, DC

3,330 abstracts, 341 of them talks: a baseline talk rate of 10.2%.

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
accessibilityalso chromatin accessibility	26	75	3.24	2.05–5.12
effector	13	23	5.19	2.63–10.23
single-cell	38	147	2.44	1.68–3.55
cell typesalso types	43	177	2.31	1.62–3.29
evolutionary	19	50	3.52	2.06–6.01
expansionsalso repeat expansions	15	33	4.19	2.27–7.73
rare variation	7	2	26.79	6.38–112.60
discovered	32	120	2.50	1.67–3.75
disease risk	28	99	2.64	1.71–4.07
currently	49	223	2.10	1.50–2.92
date	39	163	2.26	1.56–3.26
novel associations	9	13	6.30	2.73–14.56
regulatory elements	16	42	3.52	1.97–6.28
constrained	11	21	4.80	2.33–9.92
biobank we identified	6	3	16.53	4.49–60.92
systematically	19	57	3.08	1.82–5.22
chromatin	39	168	2.19	1.52–3.15
neurons	29	110	2.46	1.61–3.75
vivo	19	58	3.03	1.79–5.12
protein abundance	7	8	7.87	2.93–21.13
abundance	18	55	3.02	1.76–5.18
noncoding regions	8	12	6.07	2.52–14.62
t2d risk	8	12	6.07	2.52–14.62
ultra-rare	9	16	5.15	2.30–11.53
dose	11	24	4.21	2.07–8.56
repeat	25	93	2.50	1.59–3.93
probabilistic	8	13	5.62	2.37–13.35
apply	22	78	2.61	1.61–4.23
sc-eqtl	5	3	13.95	3.63–53.53
rare variants	32	137	2.18	1.46–3.25

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
patients	75	898	0.66	0.50–0.86
case	14	275	0.44	0.25–0.75
reported	40	544	0.60	0.43–0.85
showed	47	599	0.64	0.47–0.89
female	11	212	0.45	0.25–0.83
patient	32	436	0.61	0.42–0.89
ngsalso sequencing ngs	2	106	0.20	0.06–0.70
snp	14	238	0.51	0.30–0.88
worldwide	6	146	0.38	0.17–0.83
principal componentsalso principal	5	130	0.36	0.15–0.85
performed using	3	106	0.28	0.10–0.82
conclusions	21	303	0.59	0.38–0.93
adjusted	7	148	0.43	0.20–0.90
male	9	170	0.47	0.24–0.92
include	21	300	0.60	0.38–0.94
factors	47	559	0.70	0.51–0.96
recruited	4	109	0.35	0.14–0.91
developmental delayalso delay, developmental	6	131	0.42	0.19–0.93
regulated	1	74	0.17	0.03–0.87
aim	14	220	0.56	0.32–0.96
diagnosed	12	197	0.54	0.30–0.96
remaining	4	104	0.37	0.14–0.95
described	7	138	0.46	0.22–0.97
polymorphism	1	67	0.19	0.04–0.97
false	5	113	0.41	0.17–0.98
status	16	235	0.59	0.36–0.99
moderate	2	77	0.28	0.08–0.98
biallelic	2	76	0.28	0.08–0.99
illumina	10	168	0.53	0.28–1.00
subjects	8	145	0.50	0.25–1.00

ASHG 2024 · Denver

2,950 abstracts, 339 of them talks: a baseline talk rate of 11.5%.

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
human	141	674	2.05	1.62–2.58
chromatin accessibilityalso accessibility	25	47	4.38	2.67–7.18
plofalso plof variants	12	11	8.63	3.84–19.38
chromatin	43	139	2.60	1.81–3.73
crispr	24	54	3.64	2.23–5.95
qtlsalso trait loci, quantitative trait loci, quantitative trait, quantitative, trait	17	30	4.59	2.52–8.36
massively parallel reporteralso parallel reporter, massively parallel, massively, reporter, parallel reporter assays, mpra	12	14	6.84	3.18–14.70
neurons	34	103	2.74	1.83–4.10
atac-seq	21	47	3.65	2.16–6.15
fine-mapping	24	64	3.07	1.90–4.96
anthropometricalso anthropometric traits	9	8	8.80	3.46–22.37
africans	13	21	4.98	2.50–9.94
diverse	77	353	1.89	1.43–2.49
transcriptomes	11	16	5.51	2.57–11.79
crispri	8	7	8.90	3.31–23.93
postmortem	12	21	4.60	2.27–9.32
genetic ancestry groups	7	5	10.69	3.54–32.31
missed	19	49	3.15	1.84–5.39
ancestry groups	14	29	3.90	2.06–7.39
ukalso uk biobank	74	349	1.82	1.37–2.40
traits	80	389	1.77	1.35–2.32
asians	13	26	4.03	2.07–7.85
largest	37	138	2.21	1.51–3.24
regions	84	415	1.75	1.34–2.28
regulatory	57	251	1.91	1.40–2.61
finally	46	188	2.04	1.45–2.87
putative	28	94	2.44	1.58–3.77
cells	92	471	1.70	1.31–2.20
causal variant	11	20	4.42	2.13–9.19
consistent	53	232	1.91	1.39–2.64

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
patient	26	353	0.54	0.36–0.82
diagnosed	8	176	0.35	0.18–0.71
patients	65	694	0.66	0.50–0.87
diagnosis	32	401	0.58	0.40–0.85
suggested	2	111	0.17	0.05–0.58
clinical	83	831	0.70	0.54–0.90
based	60	637	0.67	0.50–0.90
literature	3	112	0.23	0.08–0.67
software	2	99	0.19	0.05–0.66
array	7	143	0.39	0.18–0.82
accurately	6	132	0.36	0.16–0.81
cases	54	572	0.68	0.50–0.92
exome sequencingalso exome	13	202	0.49	0.28–0.86
illumina	6	131	0.37	0.17–0.81
poor	1	81	0.14	0.03–0.69
genetic factors	4	107	0.31	0.12–0.81
control	29	347	0.62	0.42–0.92
tools	15	214	0.53	0.31–0.91
likely pathogenic	6	124	0.39	0.18–0.86
management	8	143	0.44	0.22–0.89
subjects	5	111	0.37	0.16–0.87
identification	17	228	0.57	0.34–0.93
panel	15	209	0.55	0.32–0.93
microarray	1	69	0.16	0.03–0.82
commonly	6	117	0.41	0.19–0.92
university	2	73	0.26	0.07–0.91
experience	3	82	0.32	0.11–0.94
female	10	153	0.51	0.27–0.96
use	41	433	0.70	0.50–0.98
tumor	6	111	0.44	0.20–0.97

ASHG 2025 · Boston

3,154 abstracts, 263 of them talks: a baseline talk rate of 8.3%. 16 abstracts are cut off mid-text in the booklet itself; they are left out of the phrase counts.

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
loci	101	561	2.88	2.20–3.77
million	51	232	3.00	2.15–4.19
new	76	444	2.46	1.84–3.27
regulation	62	368	2.31	1.70–3.14
non-coding	31	133	3.01	1.99–4.54
gwas	89	620	2.07	1.57–2.72
largest	32	144	2.87	1.91–4.30
identified N novel	8	8	12.04	4.61–31.44
enhancer	21	73	3.64	2.21–6.00
causal	52	301	2.31	1.66–3.20
hundreds	19	63	3.80	2.25–6.43
cell type-specificalso type-specific	15	42	4.47	2.46–8.12
motifs	16	49	4.09	2.31–7.26
novel	102	781	1.90	1.46–2.48
implicated	39	209	2.43	1.68–3.50
reporter assay mpraalso assay mpra, parallel reporter, massively parallel reporter, parallel reporter assay, mpra, reporter assay, massively parallel, reporter, massively	6	4	17.27	5.16–57.86
non-coding variants	11	26	5.26	2.60–10.64
causal genes	12	31	4.82	2.47–9.41
regulatory	61	399	2.06	1.51–2.80
cis-regulatory	16	53	3.78	2.14–6.67
genes involved	18	65	3.48	2.04–5.93
revealing	24	105	2.90	1.83–4.59
elements	31	155	2.56	1.70–3.85
risk loci	18	67	3.37	1.98–5.74
driver genes	6	6	11.95	4.00–35.72
expression	104	830	1.81	1.39–2.35
crispr	17	63	3.38	1.96–5.84
bowel disease ibdalso disease ibd, inflammatory bowel, inflammatory bowel disease, bowel, bowel disease	7	11	7.81	3.08–19.78
pinpointed	6	7	10.35	3.59–29.83
convergent	9	20	5.58	2.56–12.18

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
patients	34	672	0.53	0.37–0.77
regression	12	339	0.40	0.22–0.71
patient	12	329	0.41	0.23–0.74
pathogenic variantsalso pathogenic	6	236	0.30	0.14–0.67
typically	2	163	0.17	0.05–0.60
improve	16	365	0.49	0.30–0.82
family	9	256	0.41	0.21–0.79
syndrome	12	292	0.47	0.26–0.84
investigated	4	167	0.30	0.12–0.77
sample	21	411	0.57	0.36–0.90
features	22	420	0.59	0.38–0.92
classification	5	172	0.36	0.15–0.84
children	2	125	0.22	0.06–0.79
results	105	1451	0.73	0.57–0.95
literature	2	123	0.23	0.06–0.80
calculated	2	122	0.23	0.07–0.81
autosomal	7	193	0.43	0.21–0.91
initial	0	120	0.05	0.00–0.75
increasing	5	159	0.39	0.16–0.92
group	9	219	0.48	0.25–0.94
overlap	5	158	0.39	0.17–0.92
approach	40	641	0.68	0.48–0.97
management	5	157	0.39	0.17–0.93
age	37	595	0.69	0.48–0.98
variant	64	935	0.74	0.55–0.99
polygenicalso polygenic risk scores, risk scores, polygenic risk	18	341	0.60	0.37–0.98
developmental	12	256	0.55	0.30–0.98
symptoms	5	151	0.41	0.17–0.97
recessivealso autosomal recessive	4	135	0.38	0.15–0.97
logisticalso logistic regression	3	119	0.33	0.11–0.97

ASHG 2026 · Montreal

2,965 abstracts, 252 of them talks: a baseline talk rate of 8.5%.

More common in talks

Phrase	Talks	Posters	Odds ratio	95% CI
resource	34	151	2.67	1.80–3.96
nearly half	7	4	18.39	5.67–59.62
genetic variation	36	166	2.58	1.76–3.79
regulatory	64	390	2.04	1.50–2.75
disease loci	6	5	12.99	4.14–40.73
h²	6	5	12.99	4.14–40.73
autoimmune diseases	8	13	6.95	2.92–16.57
qtls	10	22	5.18	2.46–10.90
human	95	683	1.80	1.38–2.36
arise	12	33	4.16	2.14–8.07
enriched	57	354	1.96	1.43–2.68
additionally identified	5	4	13.38	3.82–46.82
neanderthal	5	4	13.38	3.82–46.82
antagonistic	5	4	13.38	3.82–46.82
linking	27	129	2.43	1.58–3.75
traits	61	397	1.87	1.38–2.54
rare damaging	5	5	10.94	3.33–35.94
informs	5	5	10.94	3.33–35.94
origins	7	13	6.11	2.48–15.06
opposite	9	23	4.47	2.08–9.60
nearly	18	74	2.79	1.65–4.73
exists	7	14	5.69	2.33–13.87
identified pathogenic	5	6	9.26	2.95–29.03
human genetic variation	5	6	9.26	2.95–29.03
chromatin states	5	6	9.26	2.95–29.03
kinnex	5	6	9.26	2.95–29.03
novel	76	541	1.74	1.31–2.31
large	56	365	1.85	1.35–2.53
consequences	24	116	2.39	1.51–3.77
systematic identification	5	7	8.02	2.65–24.28

More common in posters

Phrase	Talks	Posters	Odds ratio	95% CI
future	11	279	0.41	0.23–0.76
need	8	228	0.38	0.19–0.76
case	7	207	0.37	0.18–0.77
included	24	429	0.57	0.37–0.88
assess	12	271	0.47	0.26–0.84
quality	9	227	0.43	0.22–0.83
study	73	1005	0.70	0.52–0.92
highlights	3	133	0.27	0.09–0.79
normal	6	174	0.38	0.17–0.85
markers	2	116	0.22	0.06–0.78
snps	6	171	0.39	0.18–0.86
support	33	520	0.64	0.44–0.94
used	45	663	0.68	0.49–0.94
principal componentsalso principal	2	111	0.23	0.07–0.82
plasma	5	153	0.37	0.16–0.88
low	10	223	0.48	0.26–0.91
body	4	137	0.34	0.13–0.88
compared	43	625	0.69	0.49–0.97
illumina	4	130	0.36	0.14–0.93
extracted	1	86	0.18	0.04–0.91
population-specific	2	98	0.27	0.08–0.94
investigated	6	154	0.44	0.20–0.97
significantly associated	2	96	0.27	0.08–0.96
platforms	1	83	0.19	0.04–0.95
essential	11	221	0.54	0.29–0.98
suggest	20	337	0.62	0.39–0.99
methods	81	1044	0.76	0.58–1.00
conducted	22	356	0.65	0.41–1.01
conclusion	35	514	0.70	0.48–1.01
review	6	147	0.46	0.21–1.02

Phrase trends

Year by year, 2022–2026: the share of talks and of posters mentioning selected phrases. For any other phrase, and who writes about it, look it up above.

Biobanks

How often abstracts mention the most-cited biobanks, 2022–2026, counting common acronyms (UKB, MVP, BBJ). UK Biobank is in about one abstract in eight every year; All of Us grew six-fold. To compare talks with posters for any of them, look it up above.

What changed in 2026

Clinical vocabulary stopped leaning toward posters. “Patients”, “likely pathogenic” and “diagnosis” leaned toward posters over 2022–2025 (pooled odds ratios 0.6–0.8); in 2026 all three were at least as common in talks. Scanning every phrase for the largest 2026 shifts turns up more of the same: “idiopathic”, “genetic disease” and “therapy”. Click one to look it up.

Suggestive, not established. The top row was chosen before looking at 2026; the bottom row was picked for having the largest shifts, so its p-values overstate the evidence. Across all ~10,800 phrases tested, no single shift survives a false-discovery-rate correction.

Sample size

Talks report larger cohorts than posters, every year: the largest human-subject sample size each abstract states (“12,000 individuals”, “N = 450”); the bars span the middle half of abstracts that state one.

Words, A to Z

Every word in at least 10 abstracts, in alphabetical order, by the share of abstracts using it. Common English words (“the”, “of”) are left out; forms of one word (“remain”, “remains”) sit side by side and form towers, and the top of each tower is labelled. Hover a point for its word. The year buttons at the top of the page switch years; words keep their place and the scale is the same every year.

Images to download: all years · 2022 · 2023 · 2024 · 2025 · 2026

Words more common in 2026

The same words, now by how strongly each is overrepresented in 2026 abstracts compared with 2022–2025 (one-sided Fisher's exact test per word; higher is stronger). Most of the strongest are general phrasing rather than new topics: “framework” is in 27% of 2026 abstracts against 10% before, and “remains”, “incompletely”, “supporting” and “consistent” rose too. One new topic stands out: “AlphaGenome”, in 31 abstracts, none before. Hover a point for its word; click it to look it up in the table below.

−log₁₀ p from a one-sided Fisher's exact test per word; dashed line: Bonferroni correction over every word at 5%. Image to download

Every word in the figure, the most overrepresented first; scroll the table for more, click a heading to sort. Bold p: significant after Bonferroni correction. A word less common in 2026 has an odds ratio below 1 and p near 1.

Genes

How often each protein-coding gene is named, placed along the genome (hg38). APOE leads every year, with BRCA1, BRCA2 and TP53 next. Gene symbols that in abstracts mostly mean something else (MAF for minor allele frequency, HR for hazard ratio, TF for transcription factor, CAD for coronary artery disease) are left out. Hover a point for its gene; the year buttons at the top of the page switch years.

Image to download (every year, one row each)

Score your text

Paste an abstract to count how many of the 196 phrases overrepresented in ASHG talks it uses, and where that count falls among the 15,311 abstracts accepted 2022–2026.

Your text stays on your device. It is scored by this page's own script as you type; nothing is uploaded or saved. Only a rounded percentile is shown, from counts of abstracts per score.

What the score is

The phrases are the 1–3-word phrases more common in talk abstracts than in poster abstracts over 2022–2026, significant after a Bonferroni correction across all 10,816 phrases tested (z ≥ 4.58). Variants of one idea found in mostly the same abstracts (“massively parallel reporter”, “MPRA”) count once. The score is how many of them a text uses, the same way for your text as for every accepted abstract.

Mostly they mark topics (functional genomics, single-cell and chromatin work, very large cohorts) and the way large studies describe their results (“largest”, “novel”, “million”), more than style.

How well it separates talks and posters

To check the score on abstracts it was not built from, the phrase list was rebuilt from 2022–2025 alone (139 phrases) and used to score the 2,965 abstracts of 2026. A random 2026 talk outscores a random 2026 poster 64% of the time (50% would be no better than a coin flip; ties count half). For comparison, ranking abstracts by length alone (how many different words they use) puts the talk ahead 57% of the time, so the phrase count reflects more than talk abstracts running a little longer.

A description, not a verdict. Talks and posters overlap heavily at every score. Accepted abstracts only, and authors can ask for a poster, so this is association, not acceptance odds. Adding phrases to an abstract will raise its score without changing the work.