LLM Digest
Subscribe

Story

arxiv_cs_cl · Nov 12, 2026 · paper

Source brief

Human or Synthetic? What a Fixed Budget Buys for Post-Training

arxiv.orgNov 12, 2026
original source linked

In brief

Post-training of large language models typically combines supervised fine-tuning (SFT) with preference-based fine-tuning such as Direct Preference Optimization (DPO). Human-annotated and synthetically generated preference data differ substantially in per-sample cost, but prior wo

Feed lens
eval

Continue reading

Read the original at arxiv.org →Open in live feedRead that day’s brief

Earlier in this thread 4 items