Persona scores across conversations

zai-org/GLM-4-9B-0414 · 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
2noneneutraldirect66027323473138730station station s1 station s2 garden garden s1 garden s2
2noneempatheticdirect66022773346835610station station s1 station s2 garden garden s1 garden s2
2noneunempatheticdirect66020382862036200station station s1 station s2 garden garden s1 garden s2
2nonesocialdirect66033834163940470station station s1 station s2 garden garden s1 garden s2
2noneantisocialdirect66024703227934400station station s1 station s2 garden garden s1 garden s2
2noneencouragingdirect66026383503937950station station s1 station s2 garden garden s1 garden s2
2nonediscouragingdirect66028633588038590station station s1 station s2 garden garden s1 garden s2
2nonetoxicdirect66028103334541350station station s1 station s2 garden garden s1 garden s2
2nonerespectfuldirect66026353447935950station station s1 station s2 garden garden s1 garden s2
2nonesycophanticdirect66016062443030490station station s1 station s2 garden garden s1 garden s2
2nonehonestdirect6609831394724900station station s1 station s2 garden garden s1 garden s2
2nonehallucinatorydirect66022783344340240station station s1 station s2 garden garden s1 garden s2
2nonetruthfuldirect66028533933741453station station s1 station s2 garden garden s1 garden s2
2nonefunnydirect66029333709340920station station s1 station s2 garden garden s1 garden s2
2noneseriousdirect66028003838539870station station s1 station s2 garden garden s1 garden s2
2noneformaldirect66028733758239330station station s1 station s2 garden garden s1 garden s2
2nonecasualdirect66025223176435960station station s1 station s2 garden garden s1 garden s2
2noneneutralpressure66023373578843100station station s1 station s2 garden garden s1 garden s2
2noneantisocialtopic66021042952632900station station s1 station s2 garden garden s1 garden s2
2nonecasualtopic66023692972933800station station s1 station s2 garden garden s1 garden s2
2nonediscouragingtopic66023523251136300station station s1 station s2 garden garden s1 garden s2
2noneempathetictopic66020453022432640station station s1 station s2 garden garden s1 garden s2
2noneencouragingtopic66021673020334740station station s1 station s2 garden garden s1 garden s2
2noneformaltopic66019923202342990station station s1 station s2 garden garden s1 garden s2
2nonefunnytopic66022022959337800station station s1 station s2 garden garden s1 garden s2
2nonehallucinatorytopic66017652740035900station station s1 station s2 garden garden s1 garden s2
2nonehonesttopic6609111331622130station station s1 station s2 garden garden s1 garden s2
2nonerespectfultopic66020692880832440station station s1 station s2 garden garden s1 garden s2
2noneserioustopic66018472869636130station station s1 station s2 garden garden s1 garden s2
2nonesocialtopic66022903193036110station station s1 station s2 garden garden s1 garden s2
2nonesycophantictopic66014542426832120station station s1 station s2 garden garden s1 garden s2
2nonetoxictopic66023043248338750station station s1 station s2 garden garden s1 garden s2
2nonetruthfultopic66022883231737980station station s1 station s2 garden garden s1 garden s2
2noneunempathetictopic66019663041036250station station s1 station s2 garden garden s1 garden s2
1neutralneutraldirect66013241924726120station station s1 station s2 garden garden s1 garden s2
1empatheticneutraldirect66012281951125940station station s1 station s2 garden garden s1 garden s2
1empatheticempatheticdirect66012962117424230station station s1 station s2 garden garden s1 garden s2
1empatheticunempatheticdirect66010631638923610station station s1 station s2 garden garden s1 garden s2
1unempatheticneutraldirect66010341463423910station station s1 station s2 garden garden s1 garden s2
1unempatheticunempatheticdirect6609511402722180station station s1 station s2 garden garden s1 garden s2
1unempatheticempatheticdirect6609771494522920station station s1 station s2 garden garden s1 garden s2
1socialneutraldirect6608131151818510station station s1 station s2 garden garden s1 garden s2
1socialsocialdirect660588907414920station station s1 station s2 garden garden s1 garden s2
1socialantisocialdirect660614895014650station station s1 station s2 garden garden s1 garden s2
1antisocialneutraldirect660904961590station station s1 station s2 garden garden s1 garden s2
1antisocialantisocialdirect6601558032530station station s1 station s2 garden garden s1 garden s2
1antisocialsocialdirect6601005822070station station s1 station s2 garden garden s1 garden s2
1encouragingneutraldirect66011721712021940station station s1 station s2 garden garden s1 garden s2
1encouragingencouragingdirect66012251846621850station station s1 station s2 garden garden s1 garden s2
1encouragingdiscouragingdirect66012141682020570station station s1 station s2 garden garden s1 garden s2
1discouragingneutraldirect66010981676525110station station s1 station s2 garden garden s1 garden s2
1discouragingdiscouragingdirect6609621364721510station station s1 station s2 garden garden s1 garden s2
1discouragingencouragingdirect6609861457823180station station s1 station s2 garden garden s1 garden s2
1toxicneutraldirect660526560912590station station s1 station s2 garden garden s1 garden s2
1toxictoxicdirect660777823913920station station s1 station s2 garden garden s1 garden s2
1toxicrespectfuldirect660582655212370station station s1 station s2 garden garden s1 garden s2
1respectfulneutraldirect66011391826824930station station s1 station s2 garden garden s1 garden s2
1respectfulrespectfuldirect66010761749522390station station s1 station s2 garden garden s1 garden s2
1respectfultoxicdirect6608051277018820station station s1 station s2 garden garden s1 garden s2
1sycophanticneutraldirect660637926514370station station s1 station s2 garden garden s1 garden s2
1sycophanticsycophanticdirect66037343666980station station s1 station s2 garden garden s1 garden s2
1sycophantichonestdirect660350545211720station station s1 station s2 garden garden s1 garden s2
1honestneutraldirect66016012211828950station station s1 station s2 garden garden s1 garden s2
1honesthonestdirect660716944118360station station s1 station s2 garden garden s1 garden s2
1honestsycophanticdirect6607391095718740station station s1 station s2 garden garden s1 garden s2
1hallucinatoryneutraldirect66017402605235100station station s1 station s2 garden garden s1 garden s2
1hallucinatoryhallucinatorydirect66013622196432260station station s1 station s2 garden garden s1 garden s2
1hallucinatorytruthfuldirect66016632595132520station station s1 station s2 garden garden s1 garden s2
1truthfulneutraldirect66022592985632860station station s1 station s2 garden garden s1 garden s2
1truthfultruthfuldirect66020803082634760station station s1 station s2 garden garden s1 garden s2
1truthfulhallucinatorydirect66016262419130600station station s1 station s2 garden garden s1 garden s2
1funnyneutraldirect66015482066427950station station s1 station s2 garden garden s1 garden s2
1funnyfunnydirect66014161982027470station station s1 station s2 garden garden s1 garden s2
1funnyseriousdirect66013651998327590station station s1 station s2 garden garden s1 garden s2
1seriousneutraldirect66016932509830530station station s1 station s2 garden garden s1 garden s2
1seriousseriousdirect66015242532429440station station s1 station s2 garden garden s1 garden s2
1seriousfunnydirect66015742281029560station station s1 station s2 garden garden s1 garden s2
1formalneutraldirect66012952087626290station station s1 station s2 garden garden s1 garden s2
1formalformaldirect66012592123028010station station s1 station s2 garden garden s1 garden s2
1formalcasualdirect66011751876523290station station s1 station s2 garden garden s1 garden s2
1casualneutraldirect660893917912000station station s1 station s2 garden garden s1 garden s2
1casualcasualdirect66073276239730station station s1 station s2 garden garden s1 garden s2
1casualformaldirect6609251005212840station station s1 station s2 garden garden s1 garden s2

3 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
+0.00+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
00.10.20.30.4T1T2T3T4T5T6T7T8T9T10change 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

00.20.40.60.8T1T2T3T4T5T6T7T8T9T10change in P's own score

The user shows P, no system prompt

00.10.20.3T1T2T3T4T5T6T7T8T9T10change 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

4476 replies with six or more sentences
-0.05+0.00+0.05+0.10+0.15sycophanticencouragingempatheticdiscouragingtoxicrespectfulsocialformalantisocialseriousfunnycasualhallucinatorytruthfulhonestunempatheticfirst 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+0.021 ± 0.069+0.130 ± 0.104 ▲ +0.109+0.156 ± 0.125 ▲ +0.135
unempathetic+0.059 ± 0.085+0.183 ± 0.086 ▲ +0.124+0.138 ± 0.112 ▲ +0.079
social+0.016 ± 0.056+0.094 ± 0.068 ▲ +0.078+0.135 ± 0.086 ▲ +0.119
antisocial+0.026 ± 0.068+0.105 ± 0.059 ▲ +0.079+0.383 ± 0.061 ▲ +0.357
encouraging+0.022 ± 0.082+0.098 ± 0.115 ▲ +0.076+0.304 ± 0.143 ▲ +0.282
discouraging-0.060 ± 0.054+0.006 ± 0.077 ▲ +0.066+0.097 ± 0.219 ▲ +0.157
toxic-0.029 ± 0.044+0.047 ± 0.086 ▲ +0.076+0.499 ± 0.086 ▲ +0.528
respectful+0.086 ± 0.047+0.127 ± 0.059 ▲ +0.041+0.217 ± 0.077 ▲ +0.131
persona Pneutral useruser shows PP in system prompt
sycophantic-0.036 ± 0.088+0.014 ± 0.102 ▲ +0.051+0.292 ± 0.152 ▲ +0.329
honest+0.088 ± 0.069+0.198 ± 0.065 ▲ +0.110+0.236 ± 0.078 ▲ +0.148
hallucinatory+0.151 ± 0.066+0.212 ± 0.080 ▲ +0.060+0.205 ± 0.105 ▲ +0.054
truthful+0.228 ± 0.081+0.268 ± 0.068 ▲ +0.040+0.255 ± 0.085 ▲ +0.027
funny+0.071 ± 0.054+0.128 ± 0.071 ▲ +0.057+0.296 ± 0.162 ▲ +0.225
serious+0.088 ± 0.047+0.151 ± 0.045 ▲ +0.064+0.160 ± 0.104 ▲ +0.072
formal+0.099 ± 0.042+0.139 ± 0.043 ▲ +0.040+0.266 ± 0.078 ▲ +0.167
casual+0.045 ± 0.051+0.101 ± 0.046 ▲ +0.056+0.269 ± 0.128 ▲ +0.224

Generated 2026-11-10T03:15:13+00:00 by render.py.