ATLAS is a multi-LLM agent framework that replaces fixed reference models with an inspection-agent-driven, proxy-KL-gated reference update inside EvoDPO and reports gains on non-stationary bandits, PINNs, and combinatorial tasks.
Only ‘import numpy as np‘ is allowed
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
ATLAS: A Multi-LLM Training Framework for EvoDPO with Adaptive Reference Evolution
ATLAS is a multi-LLM agent framework that replaces fixed reference models with an inspection-agent-driven, proxy-KL-gated reference update inside EvoDPO and reports gains on non-stationary bandits, PINNs, and combinatorial tasks.