SSRL, a training pipeline that uses an LLM's own repeated sampling as a search environment for RL, improves question answering without external tools and transfers to real search engines.
Emergent world representations: Exploring a sequence model trained on a synthetic task
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SSRL: Self-Search Reinforcement Learning
SSRL, a training pipeline that uses an LLM's own repeated sampling as a search environment for RL, improves question answering without external tools and transfers to real search engines.