The Gradient @thegradient

**Lara J. Martin** @laramar · May 2, 2023 *

May 2, 2023 *

Yay! 2/2 papers accepted at #ACL2023 !

First, in the main conference, there's
"FIREBALL: A Dataset of Dungeons and Dragons Actual-Play with Structured Game State Information" by @zhuexe, Karmanya Aggarwal, Alex Feng, myself, & @ccb

Contributions:
- A corpus of data from people playing #DnD on Discord using a bot called #Avrae, made by @zhuexe himself. Avrae tracks vital game state information for D&D.
- #LLMs "translating" Avrae commands into plain English.

https://arxiv.org/abs/2305.01528

1/2

A diagram showing two pipelines. One has an utterance (e.g. "Filgo swings at the wolf with his axe!") and a game state ("Players & Monsters: Filgo Bitterfoot, DW1 (Dire Wolf) Attacks: Greataxe, Longsword, Shortbow"), which are fed into a LLM to produce an Avrae game command "!a great -t DW1"

Another has a dialogue history ("There's more bashing to do. Raising the head of the mace, she brings it down once again.") and a game state ("Aleksdandra attacks with a Mace! Aleksandra attacked SH1 but missed."), which are fed into a LLM to produce a narration ("The weapon comes bearing down but it is too predictable and the creature moves out of the way!")

Lara J. Martin @laramar@sigmoid.social

Next, in Findings of ACL, there's "CoRRPUS: Code-based Structured Prompting for Neurosymbolic Story Understanding" by Yijiang River Dong, myself, & @ccb

I've mentioned this work before, but now it's published! Here, we used code-based #LLMs like Codex (RIP) to provide structure in story understanding, which we saw helps the LLM figure out what characters are doing better!

https://arxiv.org/abs/2212.10754

2/2 #ACL2023