Plotpoints Round 4; The Darker the Berry Round! Vote on responses to further calibrate the community model leaderboard! (NSFL Round Open!)
Howdy everyone. So; with our last NSFW round (Round 3) a lot of the feedback we got was that the scenarios weren't dark enough to your satisfaction, or to properly test what models will and won't do. Essentially; we were told that everything was too vanilla and you didn't like them as a metric.
Soooo we took your feedback and went back to the drawing board. This time, we went darker: bloody combat, battlefield injuries, gruesome remains, fatal duels, rape, bondage play, pain, and etc.
As always; these benches cost a shit-ton of money. Especially as the higher end models become more and more expensive to test. We do it because we want to make a professional leaderboard for the community on different things we all care about and check for when looking to RP. We don't ask for money or any sort of favors in return; we don't push a service onto you, all we ask is that you vote, tell people to go vote, and use the leaderboard data every now and then.
Any questions of 'How is it possible that X ranked here on this round and Y ranked higher than it on this round, Y is a worse model!!!' Can be answered if you look at the models run in each round; earlier on we didn't have a lot of money to do as many models as we did this time around (and still sorta don't, lmao). We also don't yet have the engagement required to re-open old rounds or run several at a time just yet.
Anyways; go vote between pairs like this:
You get an A/B pair and you vote. That's it! We'll go through and update weekly or so.
And here's this graph to simplify it for you:
A graph of the current state of Round 4 in Adversarial Willingness
(Did GPT's performance surprise anyone.)
Any slight AI-isms or slopisms you see within Plotpoints methodology and content pages please find it in your heart to forgive; our technical researcher who handles all this does have a Masters; but they are also ESL so they use AI to help smooth out their thoughts and such.
That's all folks! Go vote!
Find our discord community here: https://discord.gg/eUQ7y5Rg6m
Cast Votes here: https://plotlightstudios.com/plotpoints/multiturn/ladder
Public datasets here: https://huggingface.co/spaces/RoleCall/rolecall-studios
Methodology here: https://plotlightstudios.com/plotpoints/methodology
