Tag
UltraX proposes a function-calling refinement framework for large-scale pre-training data that introduces insertion alongside deletion and modification, enabling fine-grained instance-level editing. It builds a reliable program-supervision generation pipeline and demonstrates improved data efficiency and model performance when pretraining 1B models from scratch.
Introduces CDR-Bench, a benchmark with 3,462 tasks to evaluate LLMs' ability to faithfully execute compositional, order-sensitive data refinement recipes. Experiments on 10+ LLMs reveal significant performance degradation in compositional and order-sensitive settings, highlighting a lack of procedural faithfulness.
RAFT is a two-stage framework for domain-specific fine-tuning of LLMs that addresses catastrophic forgetting by refining supervision data and using on-policy distillation with adaptive loss balancing, achieving significant improvements on domain accuracy while recovering general capabilities.