, state-action statistics) can mislabel different motions of the same strategy as diverse, failing to cover the high-level strategies that matter for coordination.
LLM-generated semantic diversity boosts zero-shot cooperation in multi-agent systems
PopGen synthesizes agent behaviors from text descriptions and improves cooperator performance against unseen partners in four collaborative environments.
Big Tech
Paul Teiletche · Kanghyun Ryu · Kyungjoon Park · Dhruva Tirumala · Markus Wulfmeier · Negar Mehr
EPFL · UC Berkeley · Yonsei University · Google DeepMind · Nomagic
Research Digest··1 min read
Thread:Social Dilemma Cooperation
Teiletche et al.
Why this paper
From Google DeepMind and 4 others
In one line
PopGen uses LLMs to generate semantically diverse agents, improving zero-shot cooperation with unseen partners.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§