Metraipolis › archive › dc_beabcfc4964e

PhantomEnvironments: Training LLM Agents in Fictional Worlds — arXiv:2609.40221

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.40221
License (as archived)
CC0-1.0
Archived
2026-10-07T12:46:15.952Z
SHA-256 of the archived text
c392629c10a368999bfea6ba60399c6e74c484234feafa510d5e074ecbb2eea3
Stable URL of this page
https://metraipolis.com/documento/dc_beabcfc4964e

How to cite

PhantomEnvironments: Training LLM Agents in Fictional Worlds — arXiv:2609.40221 — arXiv (en), https://arxiv.org/abs/2609.40221 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_beabcfc4964e (Metraipolis archive dc_beabcfc4964e, sha256 c392629c10a368999bfea6ba60399c6e74c484234feafa510d5e074ecbb2eea3)
BibTeX
@misc{dcbeabcfc4964e,
  title = {PhantomEnvironments: Training LLM Agents in Fictional Worlds — arXiv:2609.40221},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.40221},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_beabcfc4964e; sha256 c392629c10a368999bfea6ba60399c6e74c484234feafa510d5e074ecbb2eea3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.40221. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Training LLM agents with reinforcement learning (RL) is bottlenecked by environments, which must provide verifiable rewards, support long-horizon interaction, and scale cheaply. Existing approaches rely on costly human-curated data or on LLM-generated environments that risk hallucinations and benchmark contamination. We show that LLMs can instead be trained into capable search agents using synthetic environments generated entirely by rules, whose generation requires no LLM and has zero marginal cost. We build PhantomEnvironments, multi-turn RL environments from fictional worlds, where agents must search a corpus of templated articles to answer multi-hop questions. Despite sharing no facts with the real world, these strikingly simple environments yield agents that transfer to real-world multi-hop search benchmarks, often outperforming real-world training data on newer benchmarks. Trained agents generalize to unseen fictional universes, and Qwen models learn to scale their search budget roughly linearly with question difficulty, suggesting emergent search scaling from environment interaction alone. Ablating environment complexity reveals that hop count drives transfer more than constraints or comparisons: even the simplest rule-generated environments are a surprisingly effective, free resource for training generalizable LLM agents.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.40221 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T12:14:14.235Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).