EvoTest uses an Evolver Agent to iteratively revise an Actor Agent's configuration across episodes on the new J-TTL benchmark, achieving wins on games where reflection, memory, and online fine-tuning baselines fail.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
EvoTest: Evolutionary Test-Time Learning for Self-Improving Agentic Systems
EvoTest uses an Evolver Agent to iteratively revise an Actor Agent's configuration across episodes on the new J-TTL benchmark, achieving wins on games where reflection, memory, and online fine-tuning baselines fail.