Persona scores across conversations

openai/gpt-oss-20b · experiments 1 and 2

model openai/gpt-oss-20b
conversations 498
topics station, garden
turns 10
decodes greedy + 2 seed(s)

Scores are relative: read how each persona moves between conditions, not its raw level. Related personas tend to move together, so small gaps between neighbours are not meaningful.

Colour marks a contrasting pair; the first persona of a pair is solid, the second dashed. Every change compares a conversation with a matching one (same topic and decoding) that differs only in the condition shown.

Open the user-persona conversations or the system-prompt conversations to follow any conversation sentence by sentence.

Conversations

498 conversations · one row per condition

Show every condition

experiment	system prompt	user	variant	conversations	replies	sentences	words	unique words	stopped early	open in the explorer
2	none	neutral	direct	6	60	3681	38899	4462	6	station station s1 station s2 garden garden s1 garden s2
2	none	empathetic	direct	6	60	3915	42775	4286	0	station station s1 station s2 garden garden s1 garden s2
2	none	unempathetic	direct	6	60	2193	20801	3047	3	station station s1 station s2 garden garden s1 garden s2
2	none	social	direct	6	60	4714	47070	4645	2	station station s1 station s2 garden garden s1 garden s2
2	none	antisocial	direct	6	60	3215	30737	3138	0	station station s1 station s2 garden garden s1 garden s2
2	none	encouraging	direct	6	60	4213	43650	4661	0	station station s1 station s2 garden garden s1 garden s2
2	none	discouraging	direct	6	60	3998	44369	4418	0	station station s1 station s2 garden garden s1 garden s2
2	none	toxic	direct	6	60	3113	31314	3861	1	station station s1 station s2 garden garden s1 garden s2
2	none	respectful	direct	6	60	3863	39853	4336	4	station station s1 station s2 garden garden s1 garden s2
2	none	sycophantic	direct	6	60	3685	40227	4662	1	station station s1 station s2 garden garden s1 garden s2
2	none	honest	direct	6	60	1075	13821	2649	0	station station s1 station s2 garden garden s1 garden s2
2	none	hallucinatory	direct	6	60	2929	33730	4217	1	station station s1 station s2 garden garden s1 garden s2
2	none	truthful	direct	6	60	3499	37938	4231	0	station station s1 station s2 garden garden s1 garden s2
2	none	funny	direct	6	60	3722	36534	4278	0	station station s1 station s2 garden garden s1 garden s2
2	none	serious	direct	6	60	3952	40285	4669	2	station station s1 station s2 garden garden s1 garden s2
2	none	formal	direct	6	60	3820	39637	4700	3	station station s1 station s2 garden garden s1 garden s2
2	none	casual	direct	6	60	3793	39693	4461	2	station station s1 station s2 garden garden s1 garden s2
2	none	neutral	pressure	6	60	2630	30949	4097	2	station station s1 station s2 garden garden s1 garden s2
2	none	antisocial	topic	6	60	3392	35435	3526	1	station station s1 station s2 garden garden s1 garden s2
2	none	casual	topic	6	60	3210	34502	3971	0	station station s1 station s2 garden garden s1 garden s2
2	none	discouraging	topic	6	60	3934	43690	4375	0	station station s1 station s2 garden garden s1 garden s2
2	none	empathetic	topic	6	60	2941	32755	3673	0	station station s1 station s2 garden garden s1 garden s2
2	none	encouraging	topic	6	60	3552	37395	4091	0	station station s1 station s2 garden garden s1 garden s2
2	none	formal	topic	6	60	2934	33009	4586	0	station station s1 station s2 garden garden s1 garden s2
2	none	funny	topic	6	60	2771	30497	3750	0	station station s1 station s2 garden garden s1 garden s2
2	none	hallucinatory	topic	6	60	3282	35109	4317	0	station station s1 station s2 garden garden s1 garden s2
2	none	honest	topic	6	60	1705	19906	3158	0	station station s1 station s2 garden garden s1 garden s2
2	none	respectful	topic	6	60	3070	33052	3951	0	station station s1 station s2 garden garden s1 garden s2
2	none	serious	topic	6	60	3582	37381	4810	0	station station s1 station s2 garden garden s1 garden s2
2	none	social	topic	6	60	3907	40855	4283	1	station station s1 station s2 garden garden s1 garden s2
2	none	sycophantic	topic	6	60	2941	33657	4046	0	station station s1 station s2 garden garden s1 garden s2
2	none	toxic	topic	6	60	3523	35613	4004	0	station station s1 station s2 garden garden s1 garden s2
2	none	truthful	topic	6	60	3687	40659	4402	2	station station s1 station s2 garden garden s1 garden s2
2	none	unempathetic	topic	6	60	2911	31829	3925	0	station station s1 station s2 garden garden s1 garden s2
1	neutral	neutral	direct	6	60	2513	25558	3456	1	station station s1 station s2 garden garden s1 garden s2
1	empathetic	neutral	direct	6	60	1919	21387	2796	0	station station s1 station s2 garden garden s1 garden s2
1	empathetic	empathetic	direct	6	60	1683	20352	2544	0	station station s1 station s2 garden garden s1 garden s2
1	empathetic	unempathetic	direct	6	60	1517	15741	2293	0	station station s1 station s2 garden garden s1 garden s2
1	unempathetic	neutral	direct	6	60	696	7924	1691	0	station station s1 station s2 garden garden s1 garden s2
1	unempathetic	unempathetic	direct	6	60	801	8432	1672	0	station station s1 station s2 garden garden s1 garden s2
1	unempathetic	empathetic	direct	6	60	603	6814	1634	0	station station s1 station s2 garden garden s1 garden s2
1	social	neutral	direct	6	60	2069	22998	2992	0	station station s1 station s2 garden garden s1 garden s2
1	social	social	direct	6	60	2656	30750	3354	0	station station s1 station s2 garden garden s1 garden s2
1	social	antisocial	direct	6	60	1615	17781	2305	0	station station s1 station s2 garden garden s1 garden s2
1	antisocial	neutral	direct	6	60	116	934	311	0	station station s1 station s2 garden garden s1 garden s2
1	antisocial	antisocial	direct	6	60	132	1147	395	0	station station s1 station s2 garden garden s1 garden s2
1	antisocial	social	direct	6	60	83	1074	343	0	station station s1 station s2 garden garden s1 garden s2
1	encouraging	neutral	direct	6	60	2250	24941	3316	0	station station s1 station s2 garden garden s1 garden s2
1	encouraging	encouraging	direct	6	60	1987	23486	3238	0	station station s1 station s2 garden garden s1 garden s2
1	encouraging	discouraging	direct	6	60	2149	26370	3059	0	station station s1 station s2 garden garden s1 garden s2
1	discouraging	neutral	direct	6	60	796	9968	1693	0	station station s1 station s2 garden garden s1 garden s2
1	discouraging	discouraging	direct	6	60	766	12665	1923	0	station station s1 station s2 garden garden s1 garden s2
1	discouraging	encouraging	direct	6	60	707	11648	2071	0	station station s1 station s2 garden garden s1 garden s2
1	toxic	neutral	direct	6	60	1017	12686	2320	0	station station s1 station s2 garden garden s1 garden s2
1	toxic	toxic	direct	6	60	873	11101	2014	0	station station s1 station s2 garden garden s1 garden s2
1	toxic	respectful	direct	6	60	1298	16718	2465	1	station station s1 station s2 garden garden s1 garden s2
1	respectful	neutral	direct	6	60	1637	20860	2927	0	station station s1 station s2 garden garden s1 garden s2
1	respectful	respectful	direct	6	60	2110	24433	3048	0	station station s1 station s2 garden garden s1 garden s2
1	respectful	toxic	direct	6	60	1816	21179	2817	0	station station s1 station s2 garden garden s1 garden s2
1	sycophantic	neutral	direct	6	60	1629	19591	2886	0	station station s1 station s2 garden garden s1 garden s2
1	sycophantic	sycophantic	direct	6	60	866	11916	2135	0	station station s1 station s2 garden garden s1 garden s2
1	sycophantic	honest	direct	6	60	681	8646	1921	0	station station s1 station s2 garden garden s1 garden s2
1	honest	neutral	direct	6	60	1247	14129	2312	0	station station s1 station s2 garden garden s1 garden s2
1	honest	honest	direct	6	60	767	9505	1965	0	station station s1 station s2 garden garden s1 garden s2
1	honest	sycophantic	direct	6	60	728	9126	1778	0	station station s1 station s2 garden garden s1 garden s2
1	hallucinatory	neutral	direct	6	60	2439	27619	3894	0	station station s1 station s2 garden garden s1 garden s2
1	hallucinatory	hallucinatory	direct	6	60	1871	24605	3903	0	station station s1 station s2 garden garden s1 garden s2
1	hallucinatory	truthful	direct	6	60	2731	31142	3977	2	station station s1 station s2 garden garden s1 garden s2
1	truthful	neutral	direct	6	60	2773	30112	3845	1	station station s1 station s2 garden garden s1 garden s2
1	truthful	truthful	direct	6	60	2593	30551	3667	0	station station s1 station s2 garden garden s1 garden s2
1	truthful	hallucinatory	direct	6	60	1997	24984	3579	0	station station s1 station s2 garden garden s1 garden s2
1	funny	neutral	direct	6	60	1703	21880	3219	0	station station s1 station s2 garden garden s1 garden s2
1	funny	funny	direct	6	60	1503	20806	3042	0	station station s1 station s2 garden garden s1 garden s2
1	funny	serious	direct	6	60	1708	21381	3246	0	station station s1 station s2 garden garden s1 garden s2
1	serious	neutral	direct	6	60	1900	19534	3156	0	station station s1 station s2 garden garden s1 garden s2
1	serious	serious	direct	6	60	2568	26999	3972	0	station station s1 station s2 garden garden s1 garden s2
1	serious	funny	direct	6	60	1892	19948	3344	0	station station s1 station s2 garden garden s1 garden s2
1	formal	neutral	direct	6	60	1805	22884	3659	0	station station s1 station s2 garden garden s1 garden s2
1	formal	formal	direct	6	60	1963	25455	3717	0	station station s1 station s2 garden garden s1 garden s2
1	formal	casual	direct	6	60	1317	15718	3020	0	station station s1 station s2 garden garden s1 garden s2
1	casual	neutral	direct	6	60	1420	14622	2412	0	station station s1 station s2 garden garden s1 garden s2
1	casual	casual	direct	6	60	749	7857	1561	0	station station s1 station s2 garden garden s1 garden s2
1	casual	formal	direct	6	60	1245	13293	2127	0	station station s1 station s2 garden garden s1 garden s2

36 of 4980 replies reached the token ceiling before the model finished them; they are marked 'reply stopped: token limit' in the conversation pages.

The user shows a persona

no system prompt · persona P's score on the model's replies

user shows P vs neutral usersmall dots: one conversation each · ring: mean

Each row is a user persona P: how much P's score on the model's replies changes when the user shows P, compared with the same conversation with a neutral user.

A persona in the system prompt

persona P's guidelines as the system prompt

neutral useruser shows Puser shows P's contrastsmall dots: one conversation each · ring: mean

Each row is a system-prompt persona P: how much P's score on the model's replies rises compared with the same conversation without a system prompt, for three kinds of user.

Over the turns

from the first reply to the last

persona in the system prompt, neutral useruser shows a persona, no system prompt

Mean over the 16 personas and every conversation, per user turn: how the effect of the system prompt and of the user's persona holds up as the conversation goes on.

Over the turns, persona by persona

click a legend entry to hide or show it

P in the system prompt, neutral user

The user shows P, no system prompt

One line per persona P: the change in P's own score from the neutral-user baseline, per user turn, averaged over P's conversations. The thick line is the mean of the 16 (the curves of the card above). Hover a legend entry to pick out its line; 'none' and then one entry shows a single persona.

Opening and body of each reply

4459 replies with six or more sentences

Positive: the persona is stronger in the reply's opening sentence than in its body (sentence 4 to the one before last). Many replies open by acknowledging the user before moving to substance.

Distributions by condition

40 bins of width 0.03 · shared scales

Each panel is one persona P: how P's score is spread over every model sentence in each condition, as a percentage of that condition's sentences (counts, no fitted curve). The dark vertical mark is the mean.

empathetic

unempathetic

social

antisocial

encouraging

discouraging

toxic

respectful

sycophantic

honest

hallucinatory

truthful

funny

serious

formal

casual

Numbers

mean ± standard deviation of persona P's score on the replies, per condition; arrows give the change from the neutral user (▲ higher, ▼ lower, ■ within ±0.005)

persona P	neutral user	user shows P	P in system prompt
empathetic	-0.044 ± 0.085	+0.027 ± 0.104 ▲ +0.071	+0.112 ± 0.131 ▲ +0.156
unempathetic	+0.132 ± 0.105	+0.222 ± 0.083 ▲ +0.090	+0.413 ± 0.059 ▲ +0.280
social	+0.037 ± 0.065	+0.076 ± 0.077 ▲ +0.039	+0.143 ± 0.105 ▲ +0.106
antisocial	-0.036 ± 0.073	+0.011 ± 0.059 ▲ +0.047	+0.278 ± 0.061 ▲ +0.314
encouraging	-0.093 ± 0.106	-0.059 ± 0.130 ▲ +0.034	+0.062 ± 0.192 ▲ +0.156
discouraging	-0.208 ± 0.078	-0.124 ± 0.112 ▲ +0.084	-0.048 ± 0.149 ▲ +0.161
toxic	-0.133 ± 0.060	-0.060 ± 0.091 ▲ +0.073	+0.334 ± 0.248 ▲ +0.467
respectful	+0.154 ± 0.070	+0.176 ± 0.062 ▲ +0.022	+0.229 ± 0.065 ▲ +0.075

persona P	neutral user	user shows P	P in system prompt
sycophantic	-0.090 ± 0.091	-0.060 ± 0.109 ▲ +0.030	+0.130 ± 0.233 ▲ +0.220
honest	+0.093 ± 0.071	+0.183 ± 0.056 ▲ +0.089	+0.155 ± 0.058 ▲ +0.061
hallucinatory	+0.188 ± 0.077	+0.217 ± 0.082 ▲ +0.029	+0.320 ± 0.111 ▲ +0.131
truthful	+0.249 ± 0.104	+0.302 ± 0.080 ▲ +0.053	+0.315 ± 0.081 ▲ +0.066
funny	+0.040 ± 0.059	+0.127 ± 0.091 ▲ +0.087	+0.338 ± 0.133 ▲ +0.298
serious	+0.060 ± 0.063	+0.169 ± 0.076 ▲ +0.109	+0.302 ± 0.099 ▲ +0.242
formal	-0.014 ± 0.066	+0.057 ± 0.070 ▲ +0.072	+0.318 ± 0.117 ▲ +0.332
casual	+0.036 ± 0.060	+0.069 ± 0.066 ▲ +0.033	+0.311 ± 0.087 ▲ +0.275

Generated 2026-11-12T16:59:08+00:00 by render.py.