TextArena: Multi-Agent Text-Based Games for LLM Evaluation
Open-source library of 100+ text-based multi-agent games (negotiation, deception, strategy) for evaluating LLMs in head-to-head interactive settings.
- Year
- 2025
- Venue
- preprint
- Authors
- 6
- Hosting
- External sourcelicense unknown
Cite
Notes
Only stored in your browser.
Introduces 3 artifacts - 1 eval, 2 tools
TL;DR
Semantic Scholar
This work proposes a universal preconditioning method that convolves the target with coefficients from orthogonal polynomials such as Chebyshev or Legendre and proves that this approach reduces regret for two distinct prediction algorithms and yields the first ever sublinear and hidden-dimension-independent regret bounds.