A structured 4-stage pipeline harness raises task success rates to 95%+ in 2-3B parameter models while revealing format collapse and non-monotonic effects when harness support is removed.
Guidance: A guidance language for controlling large language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
It's Not the Size: Harness Design Determines Operational Stability in Small Language Models
A structured 4-stage pipeline harness raises task success rates to 95%+ in 2-3B parameter models while revealing format collapse and non-monotonic effects when harness support is removed.