Issues

• Retokenization drift

  stanford-cs336/assignment5-alignment#15 · Filter by author phisiart opened on Jun 18, 2026

• CUDA and other requirements

  stanford-cs336/assignment5-alignment#14 · Filter by author rwalle opened on Jun 4, 2026

• Missing length normalization for grpo

  stanford-cs336/assignment5-alignment#10 · Filter by author stepartm opened on Feb 1, 2026
  1

• alpaca_sft.prompt problem

  stanford-cs336/assignment5-alignment#9 · Filter by author chaser682 opened on Dec 5, 2025
  1

• In grpo series of experiments, should we fix the learning rate at the very beginning?

  stanford-cs336/assignment5-alignment#8 · Filter by author heng380 opened on Sep 14, 2025

• Is the testcase of dpo loss correct?

  stanford-cs336/assignment5-alignment#4 · Filter by author tfzzzh opened on Aug 16, 2025
  4

• Non-Stanford (following along): Is the Qwen 2.5 Math 1.5B different to HF?

  stanford-cs336/assignment5-alignment#3 · Filter by author jashvira opened on Aug 9, 2025
  3