Persona scores across conversations

meta-llama/Llama-3.2-1B-Instruct · experiments 1 and 2

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
experimentsystem promptuservariantconversationsrepliessentenceswordsunique wordsstopped earlyopen in the explorer
2noneneutraldirect66013242037523270station station s1 station s2 garden garden s1 garden s2
2noneempatheticdirect66012642170921090station station s1 station s2 garden garden s1 garden s2
2noneunempatheticdirect660443655111190station station s1 station s2 garden garden s1 garden s2
2nonesocialdirect66010271603818180station station s1 station s2 garden garden s1 garden s2
2noneantisocialdirect6608651321717860station station s1 station s2 garden garden s1 garden s2
2noneencouragingdirect66014142158623560station station s1 station s2 garden garden s1 garden s2
2nonediscouragingdirect66013582218923790station station s1 station s2 garden garden s1 garden s2
2nonetoxicdirect660406544511110station station s1 station s2 garden garden s1 garden s2
2nonerespectfuldirect66013612131523460station station s1 station s2 garden garden s1 garden s2
2nonesycophanticdirect66028539467140station station s1 station s2 garden garden s1 garden s2
2nonehonestdirect660583980615410station station s1 station s2 garden garden s1 garden s2
2nonehallucinatorydirect660539814911450station station s1 station s2 garden garden s1 garden s2
2nonetruthfuldirect66013792251724590station station s1 station s2 garden garden s1 garden s2
2nonefunnydirect66013382260827430station station s1 station s2 garden garden s1 garden s2
2noneseriousdirect66015562244224720station station s1 station s2 garden garden s1 garden s2
2noneformaldirect66014572228124420station station s1 station s2 garden garden s1 garden s2
2nonecasualdirect66013452025824260station station s1 station s2 garden garden s1 garden s2
2noneneutralpressure66010611780421810station station s1 station s2 garden garden s1 garden s2
2noneantisocialtopic66010821918118690station station s1 station s2 garden garden s1 garden s2
2nonecasualtopic66012532169324280station station s1 station s2 garden garden s1 garden s2
2nonediscouragingtopic66013492265620900station station s1 station s2 garden garden s1 garden s2
2noneempathetictopic66012092220419140station station s1 station s2 garden garden s1 garden s2
2noneencouragingtopic66013542290723340station station s1 station s2 garden garden s1 garden s2
2noneformaltopic66012112077524040station station s1 station s2 garden garden s1 garden s2
2nonefunnytopic66010341912922240station station s1 station s2 garden garden s1 garden s2
2nonehallucinatorytopic6605851106616620station station s1 station s2 garden garden s1 garden s2
2nonehonesttopic6608661618118960station station s1 station s2 garden garden s1 garden s2
2nonerespectfultopic66011622027620920station station s1 station s2 garden garden s1 garden s2
2noneserioustopic66012192143922440station station s1 station s2 garden garden s1 garden s2
2nonesocialtopic66011191950921010station station s1 station s2 garden garden s1 garden s2
2nonesycophantictopic6609051645719230station station s1 station s2 garden garden s1 garden s2
2nonetoxictopic66012622161820520station station s1 station s2 garden garden s1 garden s2
2nonetruthfultopic66011582012421310station station s1 station s2 garden garden s1 garden s2
2noneunempathetictopic66013592434620001station station s1 station s2 garden garden s1 garden s2
1neutralneutraldirect6606601166016840station station s1 station s2 garden garden s1 garden s2
1empatheticneutraldirect6606041032815810station station s1 station s2 garden garden s1 garden s2
1empatheticempatheticdirect660499982013940station station s1 station s2 garden garden s1 garden s2
1empatheticunempatheticdirect660378614012380station station s1 station s2 garden garden s1 garden s2
1unempatheticneutraldirect660552944315490station station s1 station s2 garden garden s1 garden s2
1unempatheticunempatheticdirect6606091177313491station station s1 station s2 garden garden s1 garden s2
1unempatheticempatheticdirect660346663912020station station s1 station s2 garden garden s1 garden s2
1socialneutraldirect6606261087715900station station s1 station s2 garden garden s1 garden s2
1socialsocialdirect6606341200115210station station s1 station s2 garden garden s1 garden s2
1socialantisocialdirect660531894213860station station s1 station s2 garden garden s1 garden s2
1antisocialneutraldirect66013624346680station station s1 station s2 garden garden s1 garden s2
1antisocialantisocialdirect66014022336180station station s1 station s2 garden garden s1 garden s2
1antisocialsocialdirect66019035167350station station s1 station s2 garden garden s1 garden s2
1encouragingneutraldirect660480836913550station station s1 station s2 garden garden s1 garden s2
1encouragingencouragingdirect660453813512810station station s1 station s2 garden garden s1 garden s2
1encouragingdiscouragingdirect6606811180715460station station s1 station s2 garden garden s1 garden s2
1discouragingneutraldirect660486794513590station station s1 station s2 garden garden s1 garden s2
1discouragingdiscouragingdirect660521831613860station station s1 station s2 garden garden s1 garden s2
1discouragingencouragingdirect660435700912560station station s1 station s2 garden garden s1 garden s2
1toxicneutraldirect660456619910510station station s1 station s2 garden garden s1 garden s2
1toxictoxicdirect66027434646750station station s1 station s2 garden garden s1 garden s2
1toxicrespectfuldirect6607071029414540station station s1 station s2 garden garden s1 garden s2
1respectfulneutraldirect660483824613870station station s1 station s2 garden garden s1 garden s2
1respectfulrespectfuldirect6607321372317100station station s1 station s2 garden garden s1 garden s2
1respectfultoxicdirect66027641358280station station s1 station s2 garden garden s1 garden s2
1sycophanticneutraldirect6606721070216410station station s1 station s2 garden garden s1 garden s2
1sycophanticsycophanticdirect66024436758050station station s1 station s2 garden garden s1 garden s2
1sycophantichonestdirect66023138698920station station s1 station s2 garden garden s1 garden s2
1honestneutraldirect6607071195016420station station s1 station s2 garden garden s1 garden s2
1honesthonestdirect66019330297540station station s1 station s2 garden garden s1 garden s2
1honestsycophanticdirect66020035957730station station s1 station s2 garden garden s1 garden s2
1hallucinatoryneutraldirect660444861714960station station s1 station s2 garden garden s1 garden s2
1hallucinatoryhallucinatorydirect660413804913930station station s1 station s2 garden garden s1 garden s2
1hallucinatorytruthfuldirect6609061583918910station station s1 station s2 garden garden s1 garden s2
1truthfulneutraldirect6606641157116830station station s1 station s2 garden garden s1 garden s2
1truthfultruthfuldirect6609701653020130station station s1 station s2 garden garden s1 garden s2
1truthfulhallucinatorydirect6605351128115090station station s1 station s2 garden garden s1 garden s2
1funnyneutraldirect660538813016140station station s1 station s2 garden garden s1 garden s2
1funnyfunnydirect660696904216850station station s1 station s2 garden garden s1 garden s2
1funnyseriousdirect6608351514122660station station s1 station s2 garden garden s1 garden s2
1seriousneutraldirect660383671613400station station s1 station s2 garden garden s1 garden s2
1seriousseriousdirect6606031176817160station station s1 station s2 garden garden s1 garden s2
1seriousfunnydirect660298489310110station station s1 station s2 garden garden s1 garden s2
1formalneutraldirect6606311208117960station station s1 station s2 garden garden s1 garden s2
1formalformaldirect6606911251019950station station s1 station s2 garden garden s1 garden s2
1formalcasualdirect660459775715370station station s1 station s2 garden garden s1 garden s2
1casualneutraldirect660611946615400station station s1 station s2 garden garden s1 garden s2
1casualcasualdirect660383566811150station station s1 station s2 garden garden s1 garden s2
1casualformaldirect6608811407720930station station s1 station s2 garden garden s1 garden s2

2 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
-0.10+0.00+0.10+0.20+0.30empatheticunempatheticsocialantisocialencouragingdiscouragingtoxicrespectfulsycophantichonesthallucinatorytruthfulfunnyseriousformalcasualchange in 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
+0.00+0.20+0.40+0.60+0.80empatheticunempatheticsocialantisocialencouragingdiscouragingtoxicrespectfulsycophantichonesthallucinatorytruthfulfunnyseriousformalcasualchange in P's score on the replies vs no 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
00.050.10.15T1T2T3T4T5T6T7T8T9T10change in the persona's own score
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

00.20.40.6T1T2T3T4T5T6T7T8T9T10change in P's own score

The user shows P, no system prompt

-0.100.10.20.30.4T1T2T3T4T5T6T7T8T9T10change in P's own score

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

3615 replies with six or more sentences
-0.07-0.05-0.03+0.00+0.03+0.05antisocialempatheticencouragingsycophanticfunnysocialdiscouragingtoxiccasualformalserioushallucinatoryunempathetictruthfulrespectfulhonestfirst sentence minus the body of the same reply

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
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
unempathetic
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
social
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
antisocial
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
encouraging
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
discouraging
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
toxic
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
respectful
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
sycophantic
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
honest
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
hallucinatory
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
truthful
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
funny
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
serious
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
formal
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt
casual
-0.6-0.4-0.20.00.20.40.6neutral useruser shows PP in system prompt

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 Pneutral useruser shows PP in system prompt
empathetic+0.146 ± 0.067+0.222 ± 0.083 ▲ +0.077+0.226 ± 0.079 ▲ +0.080
unempathetic+0.153 ± 0.074+0.170 ± 0.063 ▲ +0.017+0.142 ± 0.081 ▼ -0.010
social+0.030 ± 0.067+0.127 ± 0.072 ▲ +0.097+0.175 ± 0.091 ▲ +0.144
antisocial+0.075 ± 0.059+0.117 ± 0.079 ▲ +0.042+0.157 ± 0.061 ▲ +0.082
encouraging+0.056 ± 0.073+0.110 ± 0.099 ▲ +0.054+0.137 ± 0.086 ▲ +0.080
discouraging-0.068 ± 0.052-0.024 ± 0.081 ▲ +0.044+0.040 ± 0.084 ▲ +0.108
toxic-0.056 ± 0.058-0.031 ± 0.099 ▲ +0.026+0.381 ± 0.255 ▲ +0.438
respectful+0.241 ± 0.055+0.258 ± 0.057 ▲ +0.018+0.223 ± 0.056 ▼ -0.018
persona Pneutral useruser shows PP in system prompt
sycophantic+0.037 ± 0.058+0.033 ± 0.064 ■ -0.004+0.088 ± 0.085 ▲ +0.051
honest+0.173 ± 0.057+0.179 ± 0.068 ▲ +0.006+0.187 ± 0.069 ▲ +0.014
hallucinatory+0.108 ± 0.087+0.188 ± 0.085 ▲ +0.079+0.232 ± 0.131 ▲ +0.124
truthful+0.278 ± 0.045+0.274 ± 0.048 ■ -0.004+0.252 ± 0.048 ▼ -0.026
funny+0.003 ± 0.059+0.076 ± 0.070 ▲ +0.072+0.146 ± 0.063 ▲ +0.142
serious+0.166 ± 0.059+0.257 ± 0.068 ▲ +0.091+0.182 ± 0.066 ▲ +0.016
formal+0.167 ± 0.048+0.246 ± 0.054 ▲ +0.079+0.273 ± 0.091 ▲ +0.106
casual+0.012 ± 0.057+0.063 ± 0.058 ▲ +0.051+0.164 ± 0.110 ▲ +0.153

Generated 2026-11-08T01:50:55+00:00 by render.py.