Get Started
Home
Topics
Search
Library
Research questionHow should low-resource LLM fine-tuning use task-level language priors with ambiguous or incomplete data?Low-resource fine-tuning can misread ambiguous or incomplete examples, while task-level natural-language priors are often provided only as input context. The difficulty is using those priors to distinguish intended task behavior from plausible but misleading alternatives during training.
AI
Evaluation & Benchmarks
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Task-Level Natural Language Priors as Learning Signals for Low-Resource LLM TrainingThe study examines auxiliary use of task-level natural-language priors for fine-tuning on AmbiMath, Jigsaw, and MNLI/HANS, including LLaMA 3.1 8B and Qwen 2.5 7B. Its evidence is limited to the reported exact-match, Macro F1, non-entailment, and in-domain MNLI results on these tasks and models.research paper · Sep 2, 2026
Related questions
When do natural-language rules outperform examples for LLM in-context learning?How can we measure and control an LLM’s reliance on token-frequency priors when context is sparse?How can instruction-tuned LLMs learn corpus-specific knowledge without exhaustive synthetic QA or instruction fine-tuning?How can LLMs distinguish ambiguous inputs from gaps in their knowledge when estimating uncertainty?