alignment-pipeline

Tag

Cards List
#alignment-pipeline

From Context Shift to Stylistic Collapse: Why Training Objectives Matter More Than Scale

arXiv cs.CL · 2026-05-29 Cached

This paper investigates how training alignment objectives reshape linguistic features in large language models, finding that instruction-tuned systems collapse language entropy significantly more than scale would suggest, and that entropy regularization can mitigate this collapse.

0 favorites 0 likes
#alignment-pipeline

Alignment Tuning for Large Language Models: A Data-Centric Lens on Alignment Data Pipelines

arXiv cs.CL · 2026-05-27 Cached

This survey reframes the alignment tuning of large language models as a data pipeline design problem, decomposing it into three stages: response synthesis, preference evaluation, and preference instantiation. It identifies design trade-offs and failure modes, and outlines open challenges such as prompt-level alignment and agentic settings.

0 favorites 0 likes
← Back to home

Submit Feedback