An LLM-agent-driven search over fine-tuning and model-merging steps discovers post-training pipelines that beat fixed baselines by up to 9 points on tool use and 3.7 points on math, with caveats about held-out evaluation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
LaMDAgent: An Autonomous Framework for Post-Training Pipeline Optimization via LLM Agents
An LLM-agent-driven search over fine-tuning and model-merging steps discovers post-training pipelines that beat fixed baselines by up to 9 points on tool use and 3.7 points on math, with caveats about held-out evaluation.