Multi-task fine-tuning on prompted classification tasks partially generalizes to unseen domains and prompts, with identifiable failure modes mitigated by mixing with instruction tuning and unexpected benefits for thinking-based classification.
1" and no other text. If you think that the transcript does not match the criteria, respond with
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
-
How Useful Is Cross-Domain Generalization for Training LLM Monitors?
Multi-task fine-tuning on prompted classification tasks partially generalizes to unseen domains and prompts, with identifiable failure modes mitigated by mixing with instruction tuning and unexpected benefits for thinking-based classification.