Conditioning LLM training on natural-language evaluator descriptions and deploying with a held-out higher-fidelity label improved even-handedness and reduced sycophancy in two small proof-of-concept experiments.
Measuring political bias in Claude
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
method 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
method 1polarities
use method 1representative citing papers
citing papers explorer
-
Evaluation-Conditioned Training: Teaching Models to Generalize to Stronger Oversight Regimes
Conditioning LLM training on natural-language evaluator descriptions and deploying with a held-out higher-fidelity label improved even-handedness and reduced sycophancy in two small proof-of-concept experiments.